US20130293776A1 - System and method for extracting text captions from video and generating video summaries - Google Patents

System and method for extracting text captions from video and generating video summaries Download PDF

Info

Publication number
US20130293776A1
US20130293776A1 US13/935,183 US201313935183A US2013293776A1 US 20130293776 A1 US20130293776 A1 US 20130293776A1 US 201313935183 A US201313935183 A US 201313935183A US 2013293776 A1 US2013293776 A1 US 2013293776A1
Authority
US
United States
Prior art keywords
caption
frames
fields
box
video
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Abandoned
Application number
US13/935,183
Inventor
Chang Shih-Fu
Dongqing Zhang
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Columbia University of New York
Original Assignee
Columbia University of New York
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Columbia University of New York filed Critical Columbia University of New York
Priority to US13/935,183 priority Critical patent/US20130293776A1/en
Publication of US20130293776A1 publication Critical patent/US20130293776A1/en
Abandoned legal-status Critical Current

Links

Images

Classifications

    • GPHYSICS
    • G11INFORMATION STORAGE
    • G11BINFORMATION STORAGE BASED ON RELATIVE MOVEMENT BETWEEN RECORD CARRIER AND TRANSDUCER
    • G11B27/00Editing; Indexing; Addressing; Timing or synchronising; Monitoring; Measuring tape travel
    • G11B27/02Editing, e.g. varying the order of information signals recorded on, or reproduced from, record carriers
    • G11B27/031Electronic editing of digitised analogue information signals, e.g. audio or video signals
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/70Information retrieval; Database structures therefor; File system structures therefor of video data
    • G06F16/73Querying
    • G06F16/738Presentation of query results
    • G06F16/739Presentation of query results in form of a video summary, e.g. the video summary being a video sequence, a composite still image or having synthesized frames
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/70Information retrieval; Database structures therefor; File system structures therefor of video data
    • G06F16/78Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually
    • G06F16/783Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually using metadata automatically derived from the content
    • G06F16/7844Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually using metadata automatically derived from the content using original textual content or text extracted from visual content or transcript of audio data
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/70Information retrieval; Database structures therefor; File system structures therefor of video data
    • G06F16/78Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually
    • G06F16/783Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually using metadata automatically derived from the content
    • G06F16/7847Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually using metadata automatically derived from the content using low-level visual features of the video content
    • G06F16/7857Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually using metadata automatically derived from the content using low-level visual features of the video content using texture
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/70Information retrieval; Database structures therefor; File system structures therefor of video data
    • G06F16/78Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually
    • G06F16/783Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually using metadata automatically derived from the content
    • G06F16/7847Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually using metadata automatically derived from the content using low-level visual features of the video content
    • G06F16/786Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually using metadata automatically derived from the content using low-level visual features of the video content using motion, e.g. object motion or camera motion
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06TIMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00Image analysis
    • G06T7/70Determining position or orientation of objects or cameras
    • G06T7/73Determining position or orientation of objects or cameras using feature-based methods
    • G06T7/75Determining position or orientation of objects or cameras using feature-based methods involving models
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06VIMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V20/00Scenes; Scene-specific elements
    • G06V20/60Type of objects
    • G06V20/62Text, e.g. of license plates, overlay texts or captions on TV images
    • G06V20/635Overlay text, e.g. embedded captions in a TV program
    • GPHYSICS
    • G11INFORMATION STORAGE
    • G11BINFORMATION STORAGE BASED ON RELATIVE MOVEMENT BETWEEN RECORD CARRIER AND TRANSDUCER
    • G11B27/00Editing; Indexing; Addressing; Timing or synchronising; Monitoring; Measuring tape travel
    • G11B27/02Editing, e.g. varying the order of information signals recorded on, or reproduced from, record carriers
    • G11B27/031Electronic editing of digitised analogue information signals, e.g. audio or video signals
    • G11B27/034Electronic editing of digitised analogue information signals, e.g. audio or video signals on discs
    • GPHYSICS
    • G11INFORMATION STORAGE
    • G11BINFORMATION STORAGE BASED ON RELATIVE MOVEMENT BETWEEN RECORD CARRIER AND TRANSDUCER
    • G11B27/00Editing; Indexing; Addressing; Timing or synchronising; Monitoring; Measuring tape travel
    • G11B27/10Indexing; Addressing; Timing or synchronising; Measuring tape travel
    • G11B27/19Indexing; Addressing; Timing or synchronising; Measuring tape travel by using information detectable on the record carrier
    • G11B27/28Indexing; Addressing; Timing or synchronising; Measuring tape travel by using information detectable on the record carrier by using information signals recorded by the same method as the main recording
    • HELECTRICITY
    • H04ELECTRIC COMMUNICATION TECHNIQUE
    • H04NPICTORIAL COMMUNICATION, e.g. TELEVISION
    • H04N5/00Details of television systems
    • H04N5/76Television signal recording
    • H04N5/91Television signal processing therefor
    • H04N5/93Regeneration of the television signal or of selected parts thereof
    • HELECTRICITY
    • H04ELECTRIC COMMUNICATION TECHNIQUE
    • H04NPICTORIAL COMMUNICATION, e.g. TELEVISION
    • H04N7/00Television systems
    • H04N7/025Systems for the transmission of digital non-picture data, e.g. of text during the active part of a television frame
    • HELECTRICITY
    • H04ELECTRIC COMMUNICATION TECHNIQUE
    • H04NPICTORIAL COMMUNICATION, e.g. TELEVISION
    • H04N7/00Television systems
    • H04N7/08Systems for the simultaneous or sequential transmission of more than one television signal, e.g. additional information signals, the signals occupying wholly or partially the same frequency band, e.g. by time division
    • H04N7/087Systems for the simultaneous or sequential transmission of more than one television signal, e.g. additional information signals, the signals occupying wholly or partially the same frequency band, e.g. by time division with signal insertion during the vertical blanking interval only
    • H04N7/088Systems for the simultaneous or sequential transmission of more than one television signal, e.g. additional information signals, the signals occupying wholly or partially the same frequency band, e.g. by time division with signal insertion during the vertical blanking interval only the inserted signal being digital
    • H04N7/0882Systems for the simultaneous or sequential transmission of more than one television signal, e.g. additional information signals, the signals occupying wholly or partially the same frequency band, e.g. by time division with signal insertion during the vertical blanking interval only the inserted signal being digital for the transmission of character code signals, e.g. for teletext
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06VIMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V30/00Character recognition; Recognising digital ink; Document-oriented image-based pattern recognition
    • G06V30/10Character recognition
    • HELECTRICITY
    • H04ELECTRIC COMMUNICATION TECHNIQUE
    • H04NPICTORIAL COMMUNICATION, e.g. TELEVISION
    • H04N21/00Selective content distribution, e.g. interactive television or video on demand [VOD]
    • H04N21/40Client devices specifically adapted for the reception of or interaction with content, e.g. set-top-box [STB]; Operations thereof
    • H04N21/43Processing of content or additional data, e.g. demultiplexing additional data from a digital video stream; Elementary client operations, e.g. monitoring of home network or synchronising decoder's clock; Client middleware
    • H04N21/434Disassembling of a multiplex stream, e.g. demultiplexing audio and video streams, extraction of additional data from a video stream; Remultiplexing of multiplex streams; Extraction or processing of SI; Disassembling of packetised elementary stream
    • HELECTRICITY
    • H04ELECTRIC COMMUNICATION TECHNIQUE
    • H04NPICTORIAL COMMUNICATION, e.g. TELEVISION
    • H04N21/00Selective content distribution, e.g. interactive television or video on demand [VOD]
    • H04N21/40Client devices specifically adapted for the reception of or interaction with content, e.g. set-top-box [STB]; Operations thereof
    • H04N21/47End-user applications
    • H04N21/488Data services, e.g. news ticker
    • H04N21/4884Data services, e.g. news ticker for displaying subtitles

Definitions

  • the present invention relates generally to text recognition, and more particularly relates to the detection and decoding of caption regions embedded in video content and using the extracted text to generate video summaries.
  • a text caption box is embedded in the video to provide further information about the video content.
  • a video recording of a baseball game typically includes a caption box 1010 which displays game statistics such as the score, inning, ball/strike count, number of outs, etc.
  • the detection and recognition of the text captions embedded in the video frames can be an important component for video summarization, retrieval, storage and indexing. For example, by extracting a short video segment preceding certain changes in the text of the baseball caption box, such as score or number of outs, a “highlight” summary can be automatically generated.
  • Text recognition in video has been the subject of current research.
  • the article “Video OCR: Indexing Digital News Libraries by Recognition of Superimposed Captions,” by T. Sato, et al., Multimedia Systems, 7:385-394, 1999 discloses a system for detecting and recognizing text in news video. This system is described as using a line filter to enhance the text characters and a projection histogram to segment the characters. A dynamic programming algorithm is used to combine the segmentation and recognition processes to reduce the false alarms of character segmentation.
  • Past approaches to text detection in video do not adequately account for disturbances in the background areas. As a result, previous approaches are often sensitive to cluttered backgrounds, which diminish text recognition accuracy. Therefore, there remains a need for improved methods of extracting text embedded in video content. There also remains a need to improve automatic video summary generation methods using text which is extracted from the video content.
  • a method of decoding a caption box in video content is provided.
  • the expected location of a caption box in a frame of the video content is determined.
  • At least one caption box mask within the expected location is also determined.
  • a caption box mask is applied to frames of the video content and those frames exhibiting a substantial correlation to the caption box mask within the expected caption box location are identified as caption frames.
  • word regions within the confines of the expected location are identified and within each word region, text characters are identified. The text characters in the word region arc compared against a domain specific model to enhance word recognition.
  • determining an expected location of a caption box can include evaluating motion features and texture features of the video frame in the compressed domain and identifying regions having low motion features and high texture features as candidate caption box regions.
  • the method can further include evaluating the identified caption frames, within the caption box location, for changes in content; and removing caption frames from word region processing which do not exhibit a change in content.
  • a subset of the caption frames can be selected for word region processing by selecting caption frames spaced over a predetermined time period.
  • the operation of identifying text characters includes generating a vertical projection profile for each word region and identifying local inflection points, such as minima, in the vertical projection profile. Character regions can then be defined by selecting those minima which are below a threshold value as the position of character boundaries in the word region. A character recognition algorithm is then used to evaluate the defined character regions.
  • Also in accordance with the present invention is a method of generating an event based summary of video content which includes caption boxes embedded therein.
  • the summarization method begins by extracting caption boxes from at least a portion of the frames of the video content and identifying changes in the content of the extracted caption boxes which are indicative of an event of interest. For each identified change in the content of the caption box, a semantic model is applied to select a portion of the video content, preceding the change in the content of the extracted caption box, which includes the event of interest.
  • the video content is of a baseball game.
  • the semantic model can identify the portion of the video content of the event of interest as residing between a pitching event and a non-active view.
  • the pitching event can be identified using color model matching and object layout verification, such as the typical arrangement of the pitcher, batter and field.
  • Non-active view frames which generally include views of the audience or non-active players, can be identified by a reduction in the number of green pixels as compared to a preceding frame as well as a decrease in motion intensity.
  • FIG. 1 is a pictorial representation of a frame of sports video content which includes a text caption box
  • FIG. 2 is a simplified operational block diagram of an embodiment of the present system for locating a caption box in video content and extracting text therefrom;
  • FIG. 3 is a simplified flow diagram illustrating the process of determining the caption box location within the video frames
  • FIG. 4 is a pictorial representation of a block level mask used to identify frames having caption boxes located therein;
  • FIG. 5 is a pictorial representation of an exemplary average image of a caption box region
  • FIG. 6 is a pictorial representation of a text area mask derived from the average image of FIG. 5 .
  • FIG. 7 is a flow diagram further illustrating a process of identifying caption frames in video content
  • FIG. 8A is a pictorial representation of a word region extracted in accordance with the present invention.
  • FIG. 8B is a pictorial representation of the word region of FIG. 8A following intensity segmentation
  • FIG. 8C is a graph illustrating a vertical intensity profile versus pixel position for the word region of FIG. 8A ;
  • FIG. 9 is a transition graph illustrating expected state transitions in a ball-strike count encountered in video content of a baseball game
  • FIG. 10 is a pictorial diagram illustrating a frame of video with a caption box embedded therein, which is known in the prior art.
  • FIG. 11 is a temporal flow diagram illustrating a sequence of semantic events which occur in baseball video content
  • FIG. 12 is a pictorial representation of a caption box and further illustrating word region types in a baseball video domain
  • FIG. 13 is a flow diagram illustrating the process of identifying scoring events and last pitch events for generating a summary of a baseball video content from extracted caption box text;
  • FIG. 14 is a pictorial representation of an exemplary computer display of a video summary browser generated in accordance with the present invention.
  • the present invention is directed to systems and methods of locating a caption box within video content and then applying methods of optical character recognition to extract the text from the caption box. Domain knowledge can be applied to the extracted text to enhance the character recognition results. By identifying changes in the detected text which represent events of interest, the extracted caption box text can be used to form a text based and/or a video based summary of the video content.
  • video content can be represented as a sequence of discrete frames 100 through 100 -N which are displayed over time.
  • the video content can be provided in compressed form, such as using MPEG compression, as well as in high resolution pixel based form.
  • Those frames which include a caption box displayed therein are referred to herein as caption frames.
  • Those caption frames where the content within the caption box undergoes a cognizable change from the previous caption frame are referred to as keyframes.
  • FIG. 1 is a pictorial representation of a segment of video content having a number of frames 100 sequenced in time ( 100 - 1 through 100 -N).
  • a caption box region referred to as a block level bounding box 105 is identified.
  • the block level bounding box 105 is located during an initialization procedure and is used to confine processing of subsequent frames 100 - 1 through 100 -N to this portion of the video frame.
  • one or more templates will be defined to identify and process the caption box image 110 .
  • word regions 115 will be identified.
  • the word regions 115 will be further processed to define character boundaries which define character regions 120 within the word regions.
  • the pixels within the character regions 120 will preferably be subjected to character recognition processing and the word regions will be subjected to a domain specific model for word level verification, word region type identification and correction.
  • FIG. 2 is a simplified operational block diagram of a system for locating a caption bounding box 105 in the video frame 100 , identifying frames having a caption box therein, and extracting text from at least a portion of the video frames which include a caption box. It is assumed that the caption box will be consistently placed in the same region of the video frame throughout a particular piece of video content. However, because there is no predefined standard for the placement of a caption box within the flame, different video sources may place a text caption box in various locations in the video frame and may place the text within the caption box differently. Therefore, the present method begins with an initialization process which determines the location of candidate caption box bearing areas and defines masks for processing the caption boxes (step 210 ).
  • the present method for caption bounding box detection relies on two image characteristics of a caption box.
  • FIG. 3 is a flow diagram which further illustrates the operation of the initialization processing block 210 which includes determining the location of a caption bounding box 115 in a video frame 100 .
  • video content preferably in the compressed domain
  • the most common compressed domain video format follows the MPEG compression standard, such as MPEG-1, although other compression schemes can be used in the practice of the present invention.
  • the blocks are farther grouped as macro-blocks which contain four blocks (2 blocks ⁇ 2 blocks).
  • the macro-blocks include motion vectors. Motion vectors, macro-blocks and blocks are set forth in the MPEG-1 standard and are well understood in the art.
  • step 300 the motion vectors from the compressed MPEG video data are converted into a macro-block level motion energy image for the video frame 100 .
  • the motion energy image can be upscaled in width and height by the corresponding number of blocks in the macro-block width and height respectively. This upscaling operation translates the macro-block level motion energy image into the same size as a block level image.
  • caption box regions within the frame are generally static. Therefore regions exhibiting low motion energy are indicative of caption box regions.
  • the block level motion energy image can be converted to a binary motion image that indicates which regions in the frame have motion values below the threshold.
  • texture is also an indicator of text regions.
  • the discrete cosine transform (DCT) coefficients on I frames in the compressed domain MPEG video can be used to extract the texture features of the video frame (step 310 ).
  • a texture energy image is preferably generated on a block level of the MPEG video.
  • the texture features indicate how rapidly the contrast is changing in that block over successive frames.
  • text regions are generally characterized as highly textured.
  • thresholding the texture energy images can be binarized into binary texture images which distinguish between high texture regions and lower texture regions. The result is a binary texture image.
  • step 320 the binary motion energy image from step 300 and the binary texture image from step 310 are combined to form a joint motion-texture binary map.
  • the images can be combined over neighboring I and B/P frames in the MPEG video data.
  • the binary images are preferably combined using a Boolean logic AND function which removes regions that do not exhibit both low motion and high texture.
  • the binary value “1” corresponds to blocks in the compressed domain which are candidate text blocks, i.e., blocks exhibiting substantially no motion and high texture.
  • the candidate text blocks in the compressed video frame are then evaluated to form contiguous candidate caption box regions (step 330 ).
  • Many known forms of connected component analysis can be applied to form such regions.
  • a candidate text block can be selected in the motion-texture map as a seed block.
  • the neighbors of the seed block can then be evaluated to determine if they represent candidate text blocks or boundary blocks. Neighboring candidate text blocks are added to the region whereas neighboring blocks which are not candidate text blocks represent the boundary of the region which terminate region growing in that direction. The region growing process continues until each neighbor has been evaluated and either added to the region or boundary.
  • the process of forming contiguous regions is repeated throughout the video frame by selecting additional seed blocks that are not already part of a region and performing the region growing analysis until all candidate text blocks in the frame have been processed. Morphological filtering can be applied to the regions in order to remove spurious areas from the candidate regions. It will be appreciated that this form of seed based region growing is but one example of connected component analysis which can be used to identify groups of contiguous regions of candidate text blocks.
  • the candidate text regions will generally include a number of regions which do not represent likely caption box locations.
  • the candidate text regions can be evaluated over a number of frames over a time window and an incremental clustering method can be applied. For example, the candidate regions in each frame can be mapped to region clusters based on the following area overlap metric:
  • R 1 , R 2 are the two regions in consecutive frames and a(R) is the area of the region R.
  • a new region is mapped to an existing cluster lithe above metric is less than a threshold value, otherwise a new cluster is formed.
  • the clustering process stops when a dominant cluster is identified.
  • a dominant cluster is one in which at least a predetermined minimum number (such as 40%) of the frames in a continuous sliding window (e.g. 30 seconds) are mapped to the cluster.
  • the dominant cluster is used to generate a block level mask in the compressed video domain which is referred to as a Median Binary Mask (MBM) 410 , such as is illustrated in FIG. 4 .
  • MBM Median Binary Mask
  • the MBM 410 is generated by taking the median value of the member binary images forming the dominant cluster.
  • a rectangular bounding box around the MBM referred to as a Block-level Bounding Box (BBB)
  • BBB Block-level Bounding Box
  • Au example of a BBB is illustrated by bounding box 105 in FIG. 1 as well as by the outer rectangular extent 400 of the MBM 410 .
  • the MBM 410 and BBB 400 are outputs of the caption box localization process 210 which are passed to a caption frame extraction processing block 220 .
  • the BBB 105 is used to constrain processing operations in subsequent video frames ( 100 - 1 to N) to the region of the frame in which the caption box is expected to reside.
  • the MBM is applied within this region to determine if the caption box is present in a current frame. Those frames in which a caption box is present are referred to as caption frames.
  • the caption box localization and mask extraction processing block 210 also includes pixel domain processing within the BBB.
  • step 350 an averaging operation is applied to the pixels within the image constrained by the BBB, referred to as a caption image, to obtain a Representative Average Image (RAI).
  • RAI Representative Average Image
  • FIG. 5 is a pictorial example of an RAI.
  • the RAI is subjected to edge detection in order to extract the outer contour of the text area and define a Text Area Mask (TAM) 610 , an example of which is illustrated in FIG. 6 .
  • TAM Text Area Mask
  • Various forms of edge detection which are known to those skilled in the art of image processing can be used to define the TAM 610 .
  • the RAI and TAM are outputs from the initialization process that are provided from the caption box localization and mask extraction processing block 210 to a caption keyframe extraction processing block 230 .
  • the initialization processing of the caption box localization and mask extraction block 210 generally takes on the order of 30-60 seconds of video content, after which the outputs BBM, MBM, RAI, and TAM are used in a continuous process to identify caption frames and extract the caption box text from the video content data which is input into the caption frame extraction processing block 220 .
  • the video content is applied to the caption frame extraction processing block 220 and is analyzed to detect frames that are caption frames.
  • the caption frame extraction processing block 220 performs processing of the video content in the compressed domain in a portion of the video frame constrained by the BBB. Constraining the caption frame extraction processing operations allows for efficient real time processing since only a relatively small portion of the video frame is subjected to processing.
  • Step 710 the motion vectors of the compressed video blocks within the BBB are used to form a binary motion energy image.
  • a binary texture image is generated for the area within the BBB to identify blocks exhibiting high texture levels which are indicative of text (step 720 ).
  • the binary texture image and the motion energy image are combined in step 730 to generate a motion-texture binary map for the region of the current frame within the BBB.
  • Contiguous regions within the motion-texture binary map are formed using region growing or other connected component analysis techniques to identify a dominant cluster in step 740 .
  • the overlap metric of Equation 1 can then be used to compare the dominant cluster of the current frame to the MBM generated during initialization (step 750 ). If the overlap metric indicates that there is a high correlation between the dominant cluster in the current frame and that of the MBM, then the frame is identified as a caption frame in step 760 .
  • Word region detection and character recognition are processing intense operations. Thus it is desirable to minimize not only the area within a frame which is subjected to such processing but also to minimize the number of frames which are subjected to such processing.
  • caption frames are identified, it is often desirable to further identity those caption frames in which the content of the caption box changes, since these frames will generally provide the most valuable information regarding state changes in the video content.
  • Such caption frames are referred to as keyframes.
  • Keyframe extraction from the set of caption frames identified in step 220 is performed in the caption keyframe extraction block 230 .
  • Keyframe extraction is further illustrated in connection with FIG. 7 .
  • the region within the BBB is decoded from the compressed domain into the pixel domain (step 770 ).
  • the pixel domain caption image data is then compared against the RAI to eliminate false alarms. For example, a pixel-wise Euclidean distance in the RGB space between the caption image of the current frame and the RAI can be computed and compared against a predetermined threshold value (step 770 ). If in step 780 the distance exceeds the threshold value, the frame can be discarded as a false alarm (step 785 ).
  • Duplicate caption frames are identified and discarded in step 790 such that the caption frames which remain are a set of caption frames in which the content changes.
  • Duplicate caption frames can be identified by comparing the caption image of a current caption frame against the caption image of the preceding caption frame and determining the number of pixels which have changed from high intensity values to low intensity pixel values. Alternatively, comparing the caption images in consecutive frames and calculating a pixel-wise Euclidean distance in successive caption frames can be used to identify and eliminate caption frames having duplicate content. Those frames which remain will represent caption keyframes.
  • word region extraction processing and character region segmentation processing block 240 is applied to identify word regions 115 within the caption box ( FIG. 1 ).
  • Word region extraction is performed on the selected caption frames or keyframes in the pixel domain in a region confined to the TAM.
  • FIG. 8A is a pictorial representation of a word region in the pixel domain which includes the characters NY.
  • the grayscale values of the caption image are converted to a binary image by applying a suitable thresholding operation. Pixels having an intensity value higher than the threshold are considered to be text pixels.
  • Connected component analysis such as region growing, can be applied to the text pixels in the binary image to form contiguous candidate word regions. Size constraints can then be applied to the candidate regions to eliminate those regions which are either too large or too small to be considered valid word regions. Conditions can also be placed on the aspect ratio of valid word regions.
  • a temporal filtering operation can be used.
  • a temporal variance of the intensity of each pixel in the word regions is calculated using buffered caption images over a predetermined time window.
  • the resulting variance map is subjected to thresholding and region growing to identify those word regions which exhibit changes over time (such as the score or ball-strike count).
  • the temporal filtering can remove regions with static values by thresholding temporal variances of the image.
  • temporal filtering should be used with constraints that do not negatively impact such desired regions. Also, since temporal filtering requires buffering a large number of caption images, it may be too processing intense for many real time applications. In these cases, careful selection of the binarization threshold value used to identify the word regions can substantially eliminate problems arising from static graphic elements in the caption box.
  • each word region 115 is traversed in a pixel wise fashion across the width of the word region.
  • a vertical projection profile value is determined by summing the pixel values at that pixel location along the word region height.
  • FIG. 8C illustrates an example of a graph of vertical projection profile values versus pixel position for the word region of FIG. 8A .
  • the graph exhibits local minima 810 and 820 .
  • Local minimum 810 which has a value higher than threshold value 830 , represents the diagonal stroke connecting the two vertical strokes in the letter N.
  • threshold value 830 represents the diagonal stroke connecting the two vertical strokes in the letter N.
  • local minima in the vertical projection profile graph which are below a threshold value illustrated by line 830 , such as minimum 820 , represent dark regions between characters.
  • Such local minima can be used to partition the word region into individual character regions.
  • local minimum 820 establishes the location of character partition 840 between the characters N and Y in the word region illustrated in FIG. 8A . It will be appreciated that in the event of dark characters against a light background, local maxima of the vertical projection profile graph would provide the same character partition information as is provided by the local minima in FIG. 8C . Local minima and local maxima can collectively be referred to as local inflection points.
  • a pixel intensity histogram can be calculated within the word regions and this histogram used to classify the pixels within the word regions.
  • FIG. 8B illustrates the result of a three level histogram thresholding applied to the image of FIG. 8A . Those pixels exhibiting values in the highest intensity portion of the histogram can be considered character pixels 850 whereas the pixels exhibiting values in the lowest portion of the histogram can be considered background pixels 870 . Pixels exhibiting an intermediate value can be considered as “other” pixels 860 .
  • the character type pixels 850 are subjected to a character recognition processing operation.
  • a character recognition processing operation There are many known methods for performing character recognition which are suitable for use in the present system.
  • a preferred method uses Zernike moments, such as is described in “Feature Extraction Methods for Character Recognition—A Survey,” by Trier, et al., Pattern Recognition, vol. 29, pp. 641-662, 1996, the disclosure of which is hereby incorporated by reference.
  • Zernike moments are calculated based on a set of complex polynomials that form a complete orthogonal set inside the unit circle. The Zernike moments are the projection of the image onto these complex bases.
  • the character image Prior to calculating the Zernike moments, it is preferable to convert the character image into three binary images using the threshold values established in connection with the intensity histogram for each word region which identifies pixels as character, background or other pixels.
  • the first binary image distinguishes background pixels vs. other pixels.
  • the second binary image distinguishes other pixels vs. character pixels.
  • the third binary image is formed using the average value of the two thresholding values. Zernike feature vectors are then calculated for each of the three images.
  • the use of three binary images enhances the robustness of the character recognition process with respect to variations in character line width and font type.
  • each character region 120 the image of the character is projected to the complex bases to form a character feature vector.
  • the coordinate of each character pixel is normalized into the interior of the unit circle, i.e., x 2 +y 2 ⁇ 1. Since the output of the Zernike filter is a complex vector, each component of the vector includes a real and an imaginary component. The real and imaginary components can be converted to the magnitude and phase of the Zernike moments. It has been found that using nineteen (19) as the maximum order of Zernike moment and 220 as the dimension of the character feature vector provides satisfactory results.
  • Domain models can include specialized dictionaries and state transition models which represent expected state changes in the text regions.
  • various word regions can be identified by a semantic word type, such as team name, score, inning and the like, and specific models applied to the individual word regions based on the word type. For example, in a baseball video, the number of team names or abbreviations is known and can be used to identify and verify a team name word region.
  • the Zernike feature vectors for each character in the word can be concatenated to generate word-level Zernike feature vectors for the dictionary entries.
  • the Zernike character feature vectors can be concatenated to generate word-level feature vectors which can be tested against the feature vectors for the library entries. For example, a cosine distance metric can be used to compare the input word feature vector against the feature vectors for words in the dictionary which are of the same length. The closest match to the detected input word can then be selected from the dictionary.
  • certain domains can include text fields that follow an expected state transition model or heuristic.
  • a caption box generally includes a ball-strike count that must follow certain state relationships which can be used to identify and correct character recognition errors.
  • the ball-strike sequence 0-1, 0-2, 1-1 can be recognized as impermissible and flagged as an error for manual editing or corrected using further probabilistic processing.
  • a transition graph model such as illustrated in FIG. 9 , can be used as a domain specific model.
  • nodes 900 are assigned a node score and the transitions 905 between nodes assigned a transition score.
  • a node score can be defined as the best correlation value between the input word and the template word in the library.
  • the transition score is defined as the conditional probability:
  • n t , n t-1 are the nodes at the time t and the time t ⁇ 1.
  • the transition conditional probability can be estimated from observations of actual events in the domain, such as actual strike-ball sequences. A small probability is initially assigned to each conditional probability to handle possible misses or false alarms of character detection. A weighting factor ⁇ is then used to combine the node cost and transition cost as an overall cost value at one node:
  • character recognition can be performed by searching the paths of the transition graph to identify the longest paths.
  • the weighting factor ⁇ can be used to adjust the balance between the contribution from the image based text recognition methods, such as Zernike moments, and domain specific knowledge-based approach.
  • is small, the character recognition process favors the optical character recognition results.
  • the domain based model becomes more significant.
  • the extracted caption box text can be used to generate a text based abstract of the video content which indexes the state changes, such as game statistics throughout the content.
  • the caption box text can also be used as part of an event based video summarization system.
  • An event based video summarization system can use the caption box text to identify events in the video content and then select a portion of the video content based on the detected events. Recognizing that changes in certain word regions in the caption box reflect important events in the video content, a simplified summary of the content can be generated by detecting the events of interest from the caption box text and selecting a video segment of a predetermined length preceding the event. While this approach may lead to satisfactory results, use of a recognized syntactic structure of the event depicted in the video content can result in improved summary generation.
  • a summary generation system for baseball video content will now be described to illustrate the use syntactic structure of video content to refine the selection of video segments for inclusion in a summary.
  • Events in a baseball game can be classified into categories, such as score, last pitch (batter change) and base change.
  • a score event is one in which the score for either team changes.
  • a last pitch event is associated with a change in the current batter. This can be the result of a hit where the batter advances to base, a walk where the runner advances to base or an out.
  • FIG. 11 is a flow diagram which illustrates a sequence of semantic events that have been consistently observed in baseball video content.
  • the caption change 1110 which is determined from the caption box text extraction and decoding generally represents the end of an event sequence.
  • An event of interest 1120 is preceded by a pitching event 1130 .
  • a pitching event is recognizable based on the camera view which generally includes the pitcher and batter in the frame.
  • the event of interest 1120 is generally characterized by a change in camera view to one which includes a substantial view of the playing field and a resulting increase in the number of green pixels in the video frame.
  • the event 1120 is generally followed by a view of the player's dugout or the stands to illustrate reaction to the event.
  • This view referred to as a non-active view 1140 , is characterized by a marked decrease in both the motion intensity and the number of green pixels in the frame since the playing field is no longer the predominant feature in the frame.
  • a replay of the event 1150 will generally be presented followed by the change in the caption box reflecting the occurrence of the event.
  • a video segment of the event 1120 can be selected using the transition from the pitching event 1130 as the beginning of the segment and the transition to the non-active view 1140 as the end of the segment. If desired, the pitching event and a portion of the non-active view can also be included in the selected event segment to place the event in context in the resulting summary.
  • FIG. 12 illustrates an exemplary caption box for baseball video content and the word types therein.
  • Typical word types include team name 1210 , score 1220 , inning 1230 , out count 1240 and ball-strike count.
  • Domain based knowledge of the video content can be used to determine the word type for the word regions.
  • the word region type determination need be made only once during the video content since the arrangement of the caption box remains constant throughout the content.
  • the score word region 1220 is generally represented by numeric characters in the word regions in the caption box which are adjacent to the team name word regions 1210 .
  • the team name regions 1210 can be identified using a domain specific library which includes the set of team names and expected abbreviations for such names.
  • the spatial proximity and arrangement of the numeric characters also provides an indication that the numeric characters represent a score type region. For example, as illustrated in FIG. 12 , the scores for the two teams generally are vertically aligned with respect to one another.
  • the ball-strike region can be identified by the relatively frequent changes in this word region compared to other word regions as well as compliance with an expected state transition which is described above in connection with FIG. 9 .
  • a transition model recognizing that the out count word region follows the transition 0-1-2-3-0 can be used to identify such a word region.
  • summaries of baseball video content are generated using score and last pitch as events of interest. Changes in score are determined by detecting changes in the score word regions in the caption box text. Last pitch events can be detected using a decision tree model which evaluates changes in ball count and out count to identify the event.
  • FIG. 13 illustrates an event decision tree which evaluates changes in caption text to determine if a score or last pitch event have taken place.
  • the score type word region is evaluated to determine if a score change event has taken place. If the score has changed in step 1310 , than the caption change represents a score event of interest 1315 .
  • the ball-strike count region is evaluated to determine if the ball-strike count has been reset. Counts such as 0-0, 0-1 or 1-0 can be used as indicators that the ball-strike count has been reset. The use of 0-1 and 1-0 reduces the possibility of inadvertently overlooking a reset event which may arise due to character recognition errors and the like.
  • a last pitch event 1330 is detected. If the ball-strike count has been reset, a last pitch event 1330 is detected. If the ball-strike count has not been reset, the out count region is evaluated to determine if the number of outs has changed (step 1340 ). If so, than a last pitch event is detected in step 1350 , otherwise the change in caption text does not represent a last pitch 1355 or score event and is not considered an event of interest. For each caption change representing a score event or a last pitch event, a video segment of the event 1120 can be extracted as described in connection with FIG. 11 .
  • FIG. 14 illustrates an exemplary screen view of a computer-based video summary browser in accordance with the present invention.
  • the computer-based summary browser is generally implemented in software installed on a conventional computer system, such as an IBM compatible personal computer or Apple Macintosh computer system.
  • the video summary content is generally stored in computer readable media, such as a compact disk or magnetic disk storage.
  • a graphical user interface is provided which uses a digital pointer, such as a mouse, to move a selector icon on a display unit in order to control the summary browser.
  • the exemplary display of FIG. 14 provides four viewing panels which include a text window 1410 listing scoring event highlights along with a video window 1420 on which a selected scoring highlight video segment can be displayed.
  • a second text window 1430 lists last pitch events adjacent to a video window 1440 for viewing a selected last pitch event.
  • the text windows list the scoring and last pitch events of interest identified in FIG. 13 and include an activation icon 1450 for selecting the video segment associated with that event, such as can be selected using the method of FIG. 11 .
  • a user can scroll through the entries in the text window and then select and view those video segments which are of interest to the viewer by clicking on icon 1450 in a manner well known with respect to graphical user interfaces (GUI).
  • GUI graphical user interfaces
  • the specific selection and arrangement of the windows 1410 , 1420 , 1430 and 1440 in FIG. 14 is not critical to the practice of the present invention.
  • the display can be partitioned into a single text window and a single video window rather than the four windows as shown.
  • the text window and video window can take on various sizes and shapes as desired by the viewer.

Abstract

Caption boxes which are embedded in video content can be located and the text within the caption boxes decoded. Real time processing is enhanced by locating caption box regions in the compressed video domain and performing pixel based processing operations within the region of the video frame in which a caption box is located. The captions boxes are further refined by identifying word regions within the caption boxes and then applying character and word recognition processing to the identified word regions. Domain based models are used to improve text recognition results. The extracted caption box text can be used to detect events of interest in the video content and a semantic model applied to extract a segment of video of the event of interest.

Description

    FIELD OF THE INVENTION
  • The present invention relates generally to text recognition, and more particularly relates to the detection and decoding of caption regions embedded in video content and using the extracted text to generate video summaries.
  • BACKGROUND OF THE INVENTION
  • There exists a substantial volume of video and multimedia content which is available both online, such as via the Internet, and offline, such as in libraries. In such video and multimedia content, it is common for a text caption box to be embedded in the video to provide further information about the video content. For example, as illustrated in FIG. 10, a video recording of a baseball game typically includes a caption box 1010 which displays game statistics such as the score, inning, ball/strike count, number of outs, etc. The detection and recognition of the text captions embedded in the video frames can be an important component for video summarization, retrieval, storage and indexing. For example, by extracting a short video segment preceding certain changes in the text of the baseball caption box, such as score or number of outs, a “highlight” summary can be automatically generated.
  • Text recognition in video has been the subject of current research. For example, the article “Video OCR: Indexing Digital News Libraries by Recognition of Superimposed Captions,” by T. Sato, et al., Multimedia Systems, 7:385-394, 1999 discloses a system for detecting and recognizing text in news video. This system is described as using a line filter to enhance the text characters and a projection histogram to segment the characters. A dynamic programming algorithm is used to combine the segmentation and recognition processes to reduce the false alarms of character segmentation.
  • Past approaches to text detection in video do not adequately account for disturbances in the background areas. As a result, previous approaches are often sensitive to cluttered backgrounds, which diminish text recognition accuracy. Therefore, there remains a need for improved methods of extracting text embedded in video content. There also remains a need to improve automatic video summary generation methods using text which is extracted from the video content.
  • SUMMARY OF THE INVENTION
  • It is an object of the present invention to provide a system and method for location and recognition of text embedded within video content.
  • It is a further object of the present invention to provide a method of locating a caption box within video content and recognizing the text within the caption box.
  • It is another object of the present invention to provide a system and method for identifying a caption box in video content in the sports domain and detecting changes in the game state based on the text in the caption box.
  • It is yet another object of the present invention to provide a method of generating a summary of video content by detecting a caption box and selecting video segments for the summary based on changes within the caption box.
  • In accordance with the present invention, a method of decoding a caption box in video content is provided. In the method, the expected location of a caption box in a frame of the video content is determined. At least one caption box mask within the expected location is also determined. A caption box mask is applied to frames of the video content and those frames exhibiting a substantial correlation to the caption box mask within the expected caption box location are identified as caption frames. For at least a portion of the caption frames, word regions within the confines of the expected location are identified and within each word region, text characters are identified. The text characters in the word region arc compared against a domain specific model to enhance word recognition.
  • In the present method, determining an expected location of a caption box can include evaluating motion features and texture features of the video frame in the compressed domain and identifying regions having low motion features and high texture features as candidate caption box regions.
  • To enhance processing efficiency, it is desirable to remove duplicate caption frames from word region processing. Therefore, the method can further include evaluating the identified caption frames, within the caption box location, for changes in content; and removing caption frames from word region processing which do not exhibit a change in content. Alternatively, a subset of the caption frames can be selected for word region processing by selecting caption frames spaced over a predetermined time period.
  • In one embodiment, the operation of identifying text characters includes generating a vertical projection profile for each word region and identifying local inflection points, such as minima, in the vertical projection profile. Character regions can then be defined by selecting those minima which are below a threshold value as the position of character boundaries in the word region. A character recognition algorithm is then used to evaluate the defined character regions.
  • Also in accordance with the present invention is a method of generating an event based summary of video content which includes caption boxes embedded therein. The summarization method begins by extracting caption boxes from at least a portion of the frames of the video content and identifying changes in the content of the extracted caption boxes which are indicative of an event of interest. For each identified change in the content of the caption box, a semantic model is applied to select a portion of the video content, preceding the change in the content of the extracted caption box, which includes the event of interest.
  • The above described method of caption box extraction and decoding can be used in the summarization method to identify changes in the content of the extracted caption boxes
  • In one embodiment of the summarization method, the video content is of a baseball game. In this domain, the semantic model can identify the portion of the video content of the event of interest as residing between a pitching event and a non-active view. In this regard, the pitching event can be identified using color model matching and object layout verification, such as the typical arrangement of the pitcher, batter and field. Non-active view frames, which generally include views of the audience or non-active players, can be identified by a reduction in the number of green pixels as compared to a preceding frame as well as a decrease in motion intensity.
  • BRIEF DESCRIPTION OF THE DRAWING
  • Further objects, features and advantages of the invention will become apparent from the following detailed description taken in conjunction with the accompanying figures showing illustrative embodiments of the invention, in which:
  • FIG. 1 is a pictorial representation of a frame of sports video content which includes a text caption box;
  • FIG. 2 is a simplified operational block diagram of an embodiment of the present system for locating a caption box in video content and extracting text therefrom;
  • FIG. 3 is a simplified flow diagram illustrating the process of determining the caption box location within the video frames;
  • FIG. 4 is a pictorial representation of a block level mask used to identify frames having caption boxes located therein;
  • FIG. 5 is a pictorial representation of an exemplary average image of a caption box region;
  • FIG. 6 is a pictorial representation of a text area mask derived from the average image of FIG. 5.
  • FIG. 7 is a flow diagram further illustrating a process of identifying caption frames in video content;
  • FIG. 8A is a pictorial representation of a word region extracted in accordance with the present invention;
  • FIG. 8B is a pictorial representation of the word region of FIG. 8A following intensity segmentation;
  • FIG. 8C is a graph illustrating a vertical intensity profile versus pixel position for the word region of FIG. 8A;
  • FIG. 9 is a transition graph illustrating expected state transitions in a ball-strike count encountered in video content of a baseball game;
  • FIG. 10 is a pictorial diagram illustrating a frame of video with a caption box embedded therein, which is known in the prior art.
  • FIG. 11 is a temporal flow diagram illustrating a sequence of semantic events which occur in baseball video content;
  • FIG. 12 is a pictorial representation of a caption box and further illustrating word region types in a baseball video domain;
  • FIG. 13 is a flow diagram illustrating the process of identifying scoring events and last pitch events for generating a summary of a baseball video content from extracted caption box text; and
  • FIG. 14 is a pictorial representation of an exemplary computer display of a video summary browser generated in accordance with the present invention.
  • Throughout the figures, the same reference numerals and characters, unless otherwise stated, are used to denote like features, elements, components or portions of the illustrated embodiments. Moreover, while the subject invention will now be described in detail with reference to the figures, it is done so in connection with the illustrative embodiments. It is intended that changes and modifications can be made to the described embodiments without departing from the true scope and spirit of the subject invention as defined by the appended claims.
  • DETAILED DESCRIPTION OF PREFERRED EMBODIMENTS
  • The present invention is directed to systems and methods of locating a caption box within video content and then applying methods of optical character recognition to extract the text from the caption box. Domain knowledge can be applied to the extracted text to enhance the character recognition results. By identifying changes in the detected text which represent events of interest, the extracted caption box text can be used to form a text based and/or a video based summary of the video content.
  • Referring to FIG. 1, video content can be represented as a sequence of discrete frames 100 through 100-N which are displayed over time. The video content can be provided in compressed form, such as using MPEG compression, as well as in high resolution pixel based form. Those frames which include a caption box displayed therein are referred to herein as caption frames. Those caption frames where the content within the caption box undergoes a cognizable change from the previous caption frame are referred to as keyframes.
  • An overview of the operation of the present invention is provided in connection with FIG. 1, which is a pictorial representation of a segment of video content having a number of frames 100 sequenced in time (100-1 through 100-N). In frame 100, a caption box region, referred to as a block level bounding box 105 is identified. As is described in greater detail below, the block level bounding box 105 is located during an initialization procedure and is used to confine processing of subsequent frames 100-1 through 100-N to this portion of the video frame. Within the bounding box 105, one or more templates will be defined to identify and process the caption box image 110. Within the caption box image 110, word regions 115 will be identified. The word regions 115 will be further processed to define character boundaries which define character regions 120 within the word regions. The pixels within the character regions 120 will preferably be subjected to character recognition processing and the word regions will be subjected to a domain specific model for word level verification, word region type identification and correction.
  • FIG. 2 is a simplified operational block diagram of a system for locating a caption bounding box 105 in the video frame 100, identifying frames having a caption box therein, and extracting text from at least a portion of the video frames which include a caption box. It is assumed that the caption box will be consistently placed in the same region of the video frame throughout a particular piece of video content. However, because there is no predefined standard for the placement of a caption box within the flame, different video sources may place a text caption box in various locations in the video frame and may place the text within the caption box differently. Therefore, the present method begins with an initialization process which determines the location of candidate caption box bearing areas and defines masks for processing the caption boxes (step 210).
  • The present method for caption bounding box detection relies on two image characteristics of a caption box. First, the position of a caption box from frame to frame, when it is displayed, will remain essentially constant. Therefore, motion vectors for the DCT blocks in the compressed domain (MPEG) for caption box regions of the video frame will be small. Second, text is generally displayed with relatively high contrast against a background. Thus, if a text region is mapped with pixel location in the X-Y plane and pixel intensity in the Z-axis, text regions generally result in rapid and extreme changes along the Z-axis, and therefore can be considered as highly textured regions. Thus, regions within a video frame which exhibit low motion and high texture will be candidates for caption box regions.
  • FIG. 3 is a flow diagram which further illustrates the operation of the initialization processing block 210 which includes determining the location of a caption bounding box 115 in a video frame 100. Referring to FIG. 3, video content, preferably in the compressed domain, is used to identify a bounding box 115 around candidate caption box regions. The most common compressed domain video format follows the MPEG compression standard, such as MPEG-1, although other compression schemes can be used in the practice of the present invention. In MPEG-1 compressed video, pixels of the uncompressed video are represented by blocks, which represent a discrete portion of the uncompressed video (such as 1 block=8 pixel×8 pixel region). The blocks are farther grouped as macro-blocks which contain four blocks (2 blocks×2 blocks). The macro-blocks include motion vectors. Motion vectors, macro-blocks and blocks are set forth in the MPEG-1 standard and are well understood in the art.
  • In step 300, the motion vectors from the compressed MPEG video data are converted into a macro-block level motion energy image for the video frame 100. After forming the motion energy image, the motion energy image can be upscaled in width and height by the corresponding number of blocks in the macro-block width and height respectively. This upscaling operation translates the macro-block level motion energy image into the same size as a block level image. As noted above, caption box regions within the frame are generally static. Therefore regions exhibiting low motion energy are indicative of caption box regions. Using thresholding, the block level motion energy image can be converted to a binary motion image that indicates which regions in the frame have motion values below the threshold.
  • In addition to motion, texture is also an indicator of text regions. The discrete cosine transform (DCT) coefficients on I frames in the compressed domain MPEG video can be used to extract the texture features of the video frame (step 310). A texture energy image is preferably generated on a block level of the MPEG video. The texture features indicate how rapidly the contrast is changing in that block over successive frames. As noted above, text regions are generally characterized as highly textured. Using thresholding, the texture energy images can be binarized into binary texture images which distinguish between high texture regions and lower texture regions. The result is a binary texture image.
  • In step 320 the binary motion energy image from step 300 and the binary texture image from step 310 are combined to form a joint motion-texture binary map. The images can be combined over neighboring I and B/P frames in the MPEG video data. The binary images are preferably combined using a Boolean logic AND function which removes regions that do not exhibit both low motion and high texture. In the resulting joint motion-texture map, the binary value “1” corresponds to blocks in the compressed domain which are candidate text blocks, i.e., blocks exhibiting substantially no motion and high texture.
  • The candidate text blocks in the compressed video frame are then evaluated to form contiguous candidate caption box regions (step 330). Many known forms of connected component analysis can be applied to form such regions. For example, a candidate text block can be selected in the motion-texture map as a seed block. The neighbors of the seed block can then be evaluated to determine if they represent candidate text blocks or boundary blocks. Neighboring candidate text blocks are added to the region whereas neighboring blocks which are not candidate text blocks represent the boundary of the region which terminate region growing in that direction. The region growing process continues until each neighbor has been evaluated and either added to the region or boundary.
  • The process of forming contiguous regions is repeated throughout the video frame by selecting additional seed blocks that are not already part of a region and performing the region growing analysis until all candidate text blocks in the frame have been processed. Morphological filtering can be applied to the regions in order to remove spurious areas from the candidate regions. It will be appreciated that this form of seed based region growing is but one example of connected component analysis which can be used to identify groups of contiguous regions of candidate text blocks.
  • The candidate text regions will generally include a number of regions which do not represent likely caption box locations. In order to reduce the number of false alarms resulting from these candidates, the candidate text regions can be evaluated over a number of frames over a time window and an incremental clustering method can be applied. For example, the candidate regions in each frame can be mapped to region clusters based on the following area overlap metric:

  • S r(R 1 ,R 2)=1−{a(R 1 −R 1 ∩R 2)+a(R 2 −R 2 ∩R 1)}/{a(R 1)+a(R 2)}  EQ 1
  • where R1, R2 are the two regions in consecutive frames and a(R) is the area of the region R. A new region is mapped to an existing cluster lithe above metric is less than a threshold value, otherwise a new cluster is formed. The clustering process stops when a dominant cluster is identified. A dominant cluster is one in which at least a predetermined minimum number (such as 40%) of the frames in a continuous sliding window (e.g. 30 seconds) are mapped to the cluster.
  • The dominant cluster is used to generate a block level mask in the compressed video domain which is referred to as a Median Binary Mask (MBM) 410, such as is illustrated in FIG. 4. The MBM 410 is generated by taking the median value of the member binary images forming the dominant cluster. A rectangular bounding box around the MBM, referred to as a Block-level Bounding Box (BBB), is also defined (step 340). Au example of a BBB is illustrated by bounding box 105 in FIG. 1 as well as by the outer rectangular extent 400 of the MBM 410. The MBM 410 and BBB 400 are outputs of the caption box localization process 210 which are passed to a caption frame extraction processing block 220.
  • The BBB 105 is used to constrain processing operations in subsequent video frames (100-1 to N) to the region of the frame in which the caption box is expected to reside. The MBM is applied within this region to determine if the caption box is present in a current frame. Those frames in which a caption box is present are referred to as caption frames. By limiting post-initialization processing to the region within the BBB, processing efficiency is improved.
  • Returning to FIG. 3, in addition to compressed domain block level operations which generate the MBM and BBB, the caption box localization and mask extraction processing block 210 also includes pixel domain processing within the BBB. In step 350, an averaging operation is applied to the pixels within the image constrained by the BBB, referred to as a caption image, to obtain a Representative Average Image (RAI). FIG. 5 is a pictorial example of an RAI.
  • Since text pixels and text background remain generally static from frame to frame while the image pixels outside a caption box generally vary over time, the averaging operation tends to reinforce the caption box pixels while smoothing out temporal variations in the surrounding image pixels. In step 360, the RAI is subjected to edge detection in order to extract the outer contour of the text area and define a Text Area Mask (TAM) 610, an example of which is illustrated in FIG. 6. Various forms of edge detection which are known to those skilled in the art of image processing can be used to define the TAM 610. The RAI and TAM are outputs from the initialization process that are provided from the caption box localization and mask extraction processing block 210 to a caption keyframe extraction processing block 230.
  • The initialization processing of the caption box localization and mask extraction block 210 generally takes on the order of 30-60 seconds of video content, after which the outputs BBM, MBM, RAI, and TAM are used in a continuous process to identify caption frames and extract the caption box text from the video content data which is input into the caption frame extraction processing block 220.
  • Returning to FIG. 2, after initialization processing is complete in block 210, the video content is applied to the caption frame extraction processing block 220 and is analyzed to detect frames that are caption frames. The caption frame extraction processing block 220 performs processing of the video content in the compressed domain in a portion of the video frame constrained by the BBB. Constraining the caption frame extraction processing operations allows for efficient real time processing since only a relatively small portion of the video frame is subjected to processing.
  • The operation of the caption extraction processing block 220 is further described in connection with the flow chart of FIG. 7. Processing operations 710 through 760 are preferably performed in the compressed video domain. Processing is limited in each frame to the region defined by the BBB. Steps 710 through 740 are analogous to steps 300 through 330 described in connection with FIG. 3. In step 710, the motion vectors of the compressed video blocks within the BBB are used to form a binary motion energy image. In addition, a binary texture image is generated for the area within the BBB to identify blocks exhibiting high texture levels which are indicative of text (step 720). The binary texture image and the motion energy image are combined in step 730 to generate a motion-texture binary map for the region of the current frame within the BBB. Contiguous regions within the motion-texture binary map are formed using region growing or other connected component analysis techniques to identify a dominant cluster in step 740. The overlap metric of Equation 1 can then be used to compare the dominant cluster of the current frame to the MBM generated during initialization (step 750). If the overlap metric indicates that there is a high correlation between the dominant cluster in the current frame and that of the MBM, then the frame is identified as a caption frame in step 760.
  • Word region detection and character recognition are processing intense operations. Thus it is desirable to minimize not only the area within a frame which is subjected to such processing but also to minimize the number of frames which are subjected to such processing.
  • Returning to FIG. 2, after caption frames are identified, it is often desirable to further identity those caption frames in which the content of the caption box changes, since these frames will generally provide the most valuable information regarding state changes in the video content. Such caption frames are referred to as keyframes. Keyframe extraction from the set of caption frames identified in step 220 is performed in the caption keyframe extraction block 230.
  • Keyframe extraction is further illustrated in connection with FIG. 7. For a frame identified as a caption frame in step 760, the region within the BBB is decoded from the compressed domain into the pixel domain (step 770). The pixel domain caption image data is then compared against the RAI to eliminate false alarms. For example, a pixel-wise Euclidean distance in the RGB space between the caption image of the current frame and the RAI can be computed and compared against a predetermined threshold value (step 770). If in step 780 the distance exceeds the threshold value, the frame can be discarded as a false alarm (step 785). Duplicate caption frames are identified and discarded in step 790 such that the caption frames which remain are a set of caption frames in which the content changes.
  • Duplicate caption frames can be identified by comparing the caption image of a current caption frame against the caption image of the preceding caption frame and determining the number of pixels which have changed from high intensity values to low intensity pixel values. Alternatively, comparing the caption images in consecutive frames and calculating a pixel-wise Euclidean distance in successive caption frames can be used to identify and eliminate caption frames having duplicate content. Those frames which remain will represent caption keyframes.
  • As an alternative, it has been found that keyframe identification processing can be eliminated and that satisfactory results obtained by simply selecting every nth caption frame for subsequent word region extraction processing. The time between selected nth caption frames will depend on the content and the expected rate of change in the caption frames for the particular domain.
  • Returning once more to FIG. 2, after keyframes are identified or a subset of the caption frames are selected in processing block 230, word region extraction processing and character region segmentation processing block 240 is applied to identify word regions 115 within the caption box (FIG. 1). Word region extraction is performed on the selected caption frames or keyframes in the pixel domain in a region confined to the TAM. FIG. 8A is a pictorial representation of a word region in the pixel domain which includes the characters NY. To identify word regions, the grayscale values of the caption image are converted to a binary image by applying a suitable thresholding operation. Pixels having an intensity value higher than the threshold are considered to be text pixels. Connected component analysis, such as region growing, can be applied to the text pixels in the binary image to form contiguous candidate word regions. Size constraints can then be applied to the candidate regions to eliminate those regions which are either too large or too small to be considered valid word regions. Conditions can also be placed on the aspect ratio of valid word regions.
  • It has been found that graphic objects within a caption box in the proximity of the text, such as lines, boxes, logos and the like, may interfere with spatial segmentation used to identify word regions and can result in improper identification, or non-identification, of word regions in the vicinity of such objects. To correct for this potential source of error, a temporal filtering operation can be used. In one form of temporal filtering, a temporal variance of the intensity of each pixel in the word regions is calculated using buffered caption images over a predetermined time window. The resulting variance map is subjected to thresholding and region growing to identify those word regions which exhibit changes over time (such as the score or ball-strike count). The temporal filtering can remove regions with static values by thresholding temporal variances of the image. However, as certain desired fields will remain static, such as team names, temporal filtering should be used with constraints that do not negatively impact such desired regions. Also, since temporal filtering requires buffering a large number of caption images, it may be too processing intense for many real time applications. In these cases, careful selection of the binarization threshold value used to identify the word regions can substantially eliminate problems arising from static graphic elements in the caption box.
  • Referring to FIG. 1, within the word regions 115, character segmentation is applied to identify character boundaries which define character regions 120 within the word region 115. In this regard, each word region 115 is traversed in a pixel wise fashion across the width of the word region. At each pixel (column) location along the width of the word region, a vertical projection profile value is determined by summing the pixel values at that pixel location along the word region height.
  • FIG. 8C illustrates an example of a graph of vertical projection profile values versus pixel position for the word region of FIG. 8A. The graph exhibits local minima 810 and 820. Local minimum 810, which has a value higher than threshold value 830, represents the diagonal stroke connecting the two vertical strokes in the letter N. Assuming that the text characters are presented as light text against a darker background (such as white text on a black background), local minima in the vertical projection profile graph, which are below a threshold value illustrated by line 830, such as minimum 820, represent dark regions between characters. Such local minima can be used to partition the word region into individual character regions. Thus, local minimum 820 establishes the location of character partition 840 between the characters N and Y in the word region illustrated in FIG. 8A. It will be appreciated that in the event of dark characters against a light background, local maxima of the vertical projection profile graph would provide the same character partition information as is provided by the local minima in FIG. 8C. Local minima and local maxima can collectively be referred to as local inflection points.
  • In addition to evaluating the vertical projection profile of the word regions, a pixel intensity histogram can be calculated within the word regions and this histogram used to classify the pixels within the word regions. FIG. 8B illustrates the result of a three level histogram thresholding applied to the image of FIG. 8A. Those pixels exhibiting values in the highest intensity portion of the histogram can be considered character pixels 850 whereas the pixels exhibiting values in the lowest portion of the histogram can be considered background pixels 870. Pixels exhibiting an intermediate value can be considered as “other” pixels 860.
  • For each of the character regions that are identified, the character type pixels 850 are subjected to a character recognition processing operation. There are many known methods for performing character recognition which are suitable for use in the present system. A preferred method uses Zernike moments, such as is described in “Feature Extraction Methods for Character Recognition—A Survey,” by Trier, et al., Pattern Recognition, vol. 29, pp. 641-662, 1996, the disclosure of which is hereby incorporated by reference. Zernike moments are calculated based on a set of complex polynomials that form a complete orthogonal set inside the unit circle. The Zernike moments are the projection of the image onto these complex bases.
  • Prior to calculating the Zernike moments, it is preferable to convert the character image into three binary images using the threshold values established in connection with the intensity histogram for each word region which identifies pixels as character, background or other pixels. The first binary image distinguishes background pixels vs. other pixels. The second binary image distinguishes other pixels vs. character pixels. The third binary image is formed using the average value of the two thresholding values. Zernike feature vectors are then calculated for each of the three images. The use of three binary images enhances the robustness of the character recognition process with respect to variations in character line width and font type.
  • For each character region 120, the image of the character is projected to the complex bases to form a character feature vector. To calculate Zernike moments, the coordinate of each character pixel is normalized into the interior of the unit circle, i.e., x2+y2≦1. Since the output of the Zernike filter is a complex vector, each component of the vector includes a real and an imaginary component. The real and imaginary components can be converted to the magnitude and phase of the Zernike moments. It has been found that using nineteen (19) as the maximum order of Zernike moment and 220 as the dimension of the character feature vector provides satisfactory results.
  • Accuracy in the character recognition process can be improved by employing one or more domain models to verify conventional character recognition processing. Domain models can include specialized dictionaries and state transition models which represent expected state changes in the text regions. In addition, various word regions can be identified by a semantic word type, such as team name, score, inning and the like, and specific models applied to the individual word regions based on the word type. For example, in a baseball video, the number of team names or abbreviations is known and can be used to identify and verify a team name word region.
  • For each entry in the domain specific library, the Zernike feature vectors for each character in the word can be concatenated to generate word-level Zernike feature vectors for the dictionary entries. During word region recognition, for each word region identified in the caption box, the Zernike character feature vectors can be concatenated to generate word-level feature vectors which can be tested against the feature vectors for the library entries. For example, a cosine distance metric can be used to compare the input word feature vector against the feature vectors for words in the dictionary which are of the same length. The closest match to the detected input word can then be selected from the dictionary.
  • In addition to domain specific dictionary models, certain domains can include text fields that follow an expected state transition model or heuristic. For example, in a video of a baseball game, a caption box generally includes a ball-strike count that must follow certain state relationships which can be used to identify and correct character recognition errors. For example, the ball-strike sequence 0-1, 0-2, 1-1 can be recognized as impermissible and flagged as an error for manual editing or corrected using further probabilistic processing.
  • To take advantage of such specific transition rules, a transition graph model, such as illustrated in FIG. 9, can be used as a domain specific model. In the transition graph, nodes 900 are assigned a node score and the transitions 905 between nodes assigned a transition score. A node score can be defined as the best correlation value between the input word and the template word in the library. The transition score is defined as the conditional probability:

  • S t(n t-1 , n t)=p(n t |n t-1)  Eq. 2
  • where nt, nt-1 are the nodes at the time t and the time t−1. The transition conditional probability can be estimated from observations of actual events in the domain, such as actual strike-ball sequences. A small probability is initially assigned to each conditional probability to handle possible misses or false alarms of character detection. A weighting factor λ is then used to combine the node cost and transition cost as an overall cost value at one node:

  • S(n t)=λS n(n t)+(1−λ)S t(n t-1 , n t)  Eq. 3
  • Using this model, character recognition can be performed by searching the paths of the transition graph to identify the longest paths. The weighting factor λ can be used to adjust the balance between the contribution from the image based text recognition methods, such as Zernike moments, and domain specific knowledge-based approach. When λ is small, the character recognition process favors the optical character recognition results. As λ increases, the domain based model becomes more significant. The value of λ is chosen experimentally. In the baseball video domain, a value of λ=0.02 has been found to provide satisfactory results, with the knowledge based domain model improving the character recognition substantially.
  • The extracted caption box text can be used to generate a text based abstract of the video content which indexes the state changes, such as game statistics throughout the content. In addition, the caption box text can also be used as part of an event based video summarization system. An event based video summarization system can use the caption box text to identify events in the video content and then select a portion of the video content based on the detected events. Recognizing that changes in certain word regions in the caption box reflect important events in the video content, a simplified summary of the content can be generated by detecting the events of interest from the caption box text and selecting a video segment of a predetermined length preceding the event. While this approach may lead to satisfactory results, use of a recognized syntactic structure of the event depicted in the video content can result in improved summary generation.
  • A summary generation system for baseball video content will now be described to illustrate the use syntactic structure of video content to refine the selection of video segments for inclusion in a summary. Events in a baseball game can be classified into categories, such as score, last pitch (batter change) and base change. A score event is one in which the score for either team changes. A last pitch event is associated with a change in the current batter. This can be the result of a hit where the batter advances to base, a walk where the runner advances to base or an out.
  • FIG. 11 is a flow diagram which illustrates a sequence of semantic events that have been consistently observed in baseball video content. The caption change 1110 which is determined from the caption box text extraction and decoding generally represents the end of an event sequence. An event of interest 1120 is preceded by a pitching event 1130. As described in the article “Structure Analysis of Sports Video using Domain Models,” by D. Zhong and S. F. Chang, IEEE International Conference on Multimedia and Expo, Aug. 22-25, 2001, Tokyo, Japan, the disclosure of which is hereby incorporated by reference in its entirety, a pitching event is recognizable based on the camera view which generally includes the pitcher and batter in the frame.
  • The event of interest 1120 is generally characterized by a change in camera view to one which includes a substantial view of the playing field and a resulting increase in the number of green pixels in the video frame. The event 1120 is generally followed by a view of the player's dugout or the stands to illustrate reaction to the event. This view, referred to as a non-active view 1140, is characterized by a marked decrease in both the motion intensity and the number of green pixels in the frame since the playing field is no longer the predominant feature in the frame. A replay of the event 1150 will generally be presented followed by the change in the caption box reflecting the occurrence of the event. Using this syntactic model, a video segment of the event 1120 can be selected using the transition from the pitching event 1130 as the beginning of the segment and the transition to the non-active view 1140 as the end of the segment. If desired, the pitching event and a portion of the non-active view can also be included in the selected event segment to place the event in context in the resulting summary.
  • In order to determine those changes in the caption box which relate to an event of interest, semantic knowledge of the caption box text is beneficial. In this regard, a word region type determination is made for the word regions in the caption box. FIG. 12 illustrates an exemplary caption box for baseball video content and the word types therein. Typical word types include team name 1210, score 1220, inning 1230, out count 1240 and ball-strike count. Domain based knowledge of the video content can be used to determine the word type for the word regions. In general, the word region type determination need be made only once during the video content since the arrangement of the caption box remains constant throughout the content.
  • The score word region 1220 is generally represented by numeric characters in the word regions in the caption box which are adjacent to the team name word regions 1210. The team name regions 1210 can be identified using a domain specific library which includes the set of team names and expected abbreviations for such names. The spatial proximity and arrangement of the numeric characters also provides an indication that the numeric characters represent a score type region. For example, as illustrated in FIG. 12, the scores for the two teams generally are vertically aligned with respect to one another.
  • The ball-strike region can be identified by the relatively frequent changes in this word region compared to other word regions as well as compliance with an expected state transition which is described above in connection with FIG. 9. Similarly, a transition model recognizing that the out count word region follows the transition 0-1-2-3-0 can be used to identify such a word region.
  • In a preferred embodiment, summaries of baseball video content are generated using score and last pitch as events of interest. Changes in score are determined by detecting changes in the score word regions in the caption box text. Last pitch events can be detected using a decision tree model which evaluates changes in ball count and out count to identify the event.
  • FIG. 13 illustrates an event decision tree which evaluates changes in caption text to determine if a score or last pitch event have taken place. For each change in the caption box text detected in step 1305, the score type word region is evaluated to determine if a score change event has taken place. If the score has changed in step 1310, than the caption change represents a score event of interest 1315. In step 1320 the ball-strike count region is evaluated to determine if the ball-strike count has been reset. Counts such as 0-0, 0-1 or 1-0 can be used as indicators that the ball-strike count has been reset. The use of 0-1 and 1-0 reduces the possibility of inadvertently overlooking a reset event which may arise due to character recognition errors and the like. If the ball-strike count has been reset, a last pitch event 1330 is detected. If the ball-strike count has not been reset, the out count region is evaluated to determine if the number of outs has changed (step 1340). If so, than a last pitch event is detected in step 1350, otherwise the change in caption text does not represent a last pitch 1355 or score event and is not considered an event of interest. For each caption change representing a score event or a last pitch event, a video segment of the event 1120 can be extracted as described in connection with FIG. 11.
  • FIG. 14 illustrates an exemplary screen view of a computer-based video summary browser in accordance with the present invention. The computer-based summary browser is generally implemented in software installed on a conventional computer system, such as an IBM compatible personal computer or Apple Macintosh computer system. In such a system, the video summary content is generally stored in computer readable media, such as a compact disk or magnetic disk storage. A graphical user interface is provided which uses a digital pointer, such as a mouse, to move a selector icon on a display unit in order to control the summary browser.
  • The exemplary display of FIG. 14 provides four viewing panels which include a text window 1410 listing scoring event highlights along with a video window 1420 on which a selected scoring highlight video segment can be displayed. A second text window 1430 lists last pitch events adjacent to a video window 1440 for viewing a selected last pitch event. The text windows list the scoring and last pitch events of interest identified in FIG. 13 and include an activation icon 1450 for selecting the video segment associated with that event, such as can be selected using the method of FIG. 11. A user can scroll through the entries in the text window and then select and view those video segments which are of interest to the viewer by clicking on icon 1450 in a manner well known with respect to graphical user interfaces (GUI).
  • The specific selection and arrangement of the windows 1410, 1420, 1430 and 1440 in FIG. 14 is not critical to the practice of the present invention. For example, the display can be partitioned into a single text window and a single video window rather than the four windows as shown. In addition, the text window and video window can take on various sizes and shapes as desired by the viewer.
  • Although the present invention has been described in connection with specific exemplary embodiments, it should be understood that various changes, substitutions and alterations can be made to the disclosed embodiments without departing from the spirit and scope of the invention as set forth in the appended claims.

Claims (23)

1. A system for generating an event based summary of one or more fields or frames of video content including caption boxes embedded therein, comprising:
extracting means for extracting caption boxes from at least a portion of the fields or frames of the video content;
change identifying means, coupled to the extracting means and receiving the caption boxes therefrom, for identifying one or more changes in the content of the extracted caption boxes which are indicative of an event of interest; and
model applying means, coupled to the extracting means and change identifying means and receiving the caption boxes and the one or more changes therefrom, for applying a semantic model to select a portion of the video content preceding each of the one or more changes.
2. The system of claim 1, wherein the video content comprises video of a baseball game, and wherein the semantic model comprises a model which identifies the portion of the video content of the event of interest as residing between a pitching event and non-active views.
3. The system of claim 2, wherein the events of interest comprise score events and last pitch events.
4. The system of claim 1, wherein the extracting means further comprises:
location means for determining at least one expected location of a caption box in one or more fields or frames of the video content;
determining means, coupled to the location means and receiving the at least one expected location therefrom, for determining at least one caption box mask within the expected location;
frame identifying means, coupled to the determining means and location means and receiving the at least one caption box mask and the at least one expected location therefrom, for identifying one or more fields or frames in the video content as caption fields or frames if the current field or frame exhibits substantial correlation to the at least one caption box mask within the expected caption box;
word region identifying means, coupled to the frame identifying means and receiving the identified caption fields or frames therefrom, for at least a portion of the caption fields or frames, identifying word regions within the confines of the expected location; and
text character means, coupled to the word region identifying means and receiving the word regions therefrom, for each word region, identifying text characters within the region; and processing the identified text characters.
5. The system of claim 4, further comprising comparing means, coupled to the text character means and receiving the text characters therefrom, for comparing the text characters in the word region against a domain specific model to enhance word recognition.
6. The system of claim 4, wherein the location means includes means for evaluating motion features of the video field or frame in the compressed domain, evaluating texture features of the video field or frame in the compressed domain, and identifying regions having low motion features and high texture features as candidate caption box regions.
7. The system of claim 5, wherein the location means includes means for evaluating motion features of the video field or frame in the compressed domain, evaluating texture features of the video field or frame in the compressed domain, and identifying regions having low motion features and high texture features as candidate caption box regions.
8. The system of claim 4, further comprising removal means, coupled to the frame identifying means, the determining means and the location means and receiving the identified caption fields or frames, the at least one caption box mask and the at least one expected location therefrom, for evaluating the identified caption fields or frames, within the caption box location, for changes in content and removing caption fields or frames from word region processing which do not exhibit a change in content.
9. The system of claim 4, further comprising interval means, coupled to the frame identifying means and receiving the identified caption fields or frames therefrom, for selecting a subset of caption fields or frames based on a predetermined time interval and sending that subset to the word region identifying means.
10. The system of claim 4, further comprising number means, coupled to the frame identifying means and receiving the identified caption fields or frames therefrom, for determining a subset of caption fields or frames by selecting caption fields or frames based on a predetermined number of intervening caption fields or frames and sending that subset to the word region identifying means.
11. A non-transitory computer-readable storage medium storing a program for causing a computer to implement a method for generating an event based summary of one or more fields or frames of video content including caption boxes embedded therein, comprising:
extracting caption boxes from at least a portion of the fields or frames of the video content;
identifying one or more changes in the content of the extracted caption boxes which are indicative of an event of interest; and
applying a semantic model to select a portion of the video content preceding each of the one or more changes.
12. The non-transitory computer-readable storage medium of claim 11, wherein the extracting caption boxes further comprises:
determining at least one expected location of a caption box in one or more fields or frames of the video content;
determining at least one caption box mask within the expected location;
identifying one or more fields or frames in the video content as caption fields or frames if the current field or frame exhibits substantial correlation to the at least one caption box mask within the expected caption box location;
for at least a portion of the caption fields or frames, identifying word regions within the confines of the expected location;
for each word region, identifying text characters within the region, and processing the identified text characters.
13. The non-transitory computer-readable storage medium of claim 12, wherein the extracting caption boxes includes the step of comparing the text characters in the word region against a domain specific model to enhance word recognition.
14. A system for generating an event based summary of one or more fields or frames of video content including caption boxes embedded therein, comprising:
a caption box extractor for extracting caption boxes from at least a portion of the fields or frames of the video content;
a caption change identification processor, coupled to the caption box extractor and receiving the caption boxes therefrom, for identifying one or more changes in the content of the extracted caption boxes which are indicative of an event of interest; and
a model processor, coupled to the caption box extractor and caption change identification processor and receiving the caption boxes and the one or more changes therefrom, for applying a semantic model to select a portion of the video content preceding each of the one or more changes.
15. The system of claim 14, wherein the video content comprises video of a baseball game, and wherein the semantic model comprises a model which identifies the portion of the video content of the event of interest as residing between a pitching event and non-active views.
16. The system of claim 14, wherein the event of interest comprises one or more score events and last pitch events.
17. The system of claim 14, wherein the caption box extractor further comprises:
a location processor for determining at least one expected location of a caption box in one or more fields or frames of the video content;
a caption mask determining processor, coupled to the location processor and receiving the at least one expected location therefrom, for determining at least one caption box mask within the expected location;
a frame identification processor, coupled to the caption mask determining processor and location processor and receiving the at least one caption box mask and the at least one expected location therefrom, for identifying one or more fields or frames in the video content as caption fields or frames if the current field or frame exhibits substantial correlation to the at least one caption box mask within the expected caption box;
a word region identification processor, coupled to the frame identification processor and receiving the identified caption fields or frames therefrom, for at least a portion of the caption fields or frames, identifying word regions within the confines of the expected location; and
a text character identification processor, coupled to the word region identification processor and receiving the word regions therefrom, for each word region, identifying text characters within the region; and processing the identified text characters.
18. The system of claim 17, further comprising a text character comparison processor, coupled to the text character identification processor and receiving the text characters therefrom, for comparing the text characters in the word region against a domain specific model to enhance word recognition.
19. The system of claim 17, wherein the location processor includes a processor for evaluating motion features of the video field or frame in the compressed domain, evaluating texture features of the video field or frame in the compressed domain, and identifying regions having low motion features and high texture features as candidate caption box regions.
20. The system of claim 18, wherein the location processor includes a processor for evaluating motion features of the video field or frame in the compressed domain, evaluating texture features of the video field or frame in the compressed domain, and identifying regions having low motion features and high texture features as candidate caption box regions.
21. The system of claim 17, further comprising a caption removal processor, coupled to the frame identification processor, the caption mask determining processor and the location processor and receiving the identified caption fields or frames, the at least one caption box mask and the at least one expected location therefrom, for evaluating the identified caption fields or frames, within the caption box location, for changes in content and removing caption fields or frames from word region processing which do not exhibit a change in content.
22. The system of claim 17, further comprising an interval-based caption selection processor, coupled to the frame identification processor and receiving the identified caption fields or frames therefrom, for selecting a subset of caption fields or frames based on a predetermined time interval and sending that subset to the word region identification processor.
23. The system of claim 17, further comprising a number-based caption selection processor, coupled to the frame identification processor and receiving the identified caption fields or frames therefrom, for determining a subset of caption fields or frames by selecting caption fields or frames based on a predetermined number of intervening caption fields or frames and sending that subset to the word region identification processor.
US13/935,183 2001-12-06 2013-07-03 System and method for extracting text captions from video and generating video summaries Abandoned US20130293776A1 (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
US13/935,183 US20130293776A1 (en) 2001-12-06 2013-07-03 System and method for extracting text captions from video and generating video summaries

Applications Claiming Priority (5)

Application Number Priority Date Filing Date Title
US33791101P 2001-12-06 2001-12-06
PCT/US2002/039247 WO2003051031A2 (en) 2001-12-06 2002-12-06 Method and apparatus for planarization of a material by growing and removing a sacrificial film
US10/494,739 US7339992B2 (en) 2001-12-06 2002-12-06 System and method for extracting text captions from video and generating video summaries
US11/960,424 US8488682B2 (en) 2001-12-06 2007-12-19 System and method for extracting text captions from video and generating video summaries
US13/935,183 US20130293776A1 (en) 2001-12-06 2013-07-03 System and method for extracting text captions from video and generating video summaries

Related Parent Applications (1)

Application Number Title Priority Date Filing Date
US11/960,424 Division US8488682B2 (en) 2001-12-06 2007-12-19 System and method for extracting text captions from video and generating video summaries

Publications (1)

Publication Number Publication Date
US20130293776A1 true US20130293776A1 (en) 2013-11-07

Family

ID=23322541

Family Applications (3)

Application Number Title Priority Date Filing Date
US10/494,739 Expired - Fee Related US7339992B2 (en) 2001-12-06 2002-12-06 System and method for extracting text captions from video and generating video summaries
US11/960,424 Expired - Fee Related US8488682B2 (en) 2001-12-06 2007-12-19 System and method for extracting text captions from video and generating video summaries
US13/935,183 Abandoned US20130293776A1 (en) 2001-12-06 2013-07-03 System and method for extracting text captions from video and generating video summaries

Family Applications Before (2)

Application Number Title Priority Date Filing Date
US10/494,739 Expired - Fee Related US7339992B2 (en) 2001-12-06 2002-12-06 System and method for extracting text captions from video and generating video summaries
US11/960,424 Expired - Fee Related US8488682B2 (en) 2001-12-06 2007-12-19 System and method for extracting text captions from video and generating video summaries

Country Status (3)

Country Link
US (3) US7339992B2 (en)
AU (1) AU2002351310A1 (en)
WO (1) WO2003051031A2 (en)

Cited By (18)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20140059039A1 (en) * 2012-08-22 2014-02-27 Kirk Baker Methods and systems for searching and displaying a plurality of entities within an interactive user interface
WO2015126830A1 (en) * 2014-02-21 2015-08-27 Liveclips Llc System for annotating media content for automatic content understanding
US9141859B2 (en) 2008-11-17 2015-09-22 Liveclips Llc Method and system for segmenting and transmitting on-demand live-action video in real-time
US9141860B2 (en) 2008-11-17 2015-09-22 Liveclips Llc Method and system for segmenting and transmitting on-demand live-action video in real-time
WO2016036460A1 (en) * 2014-09-04 2016-03-10 Intel Corporation Real time video summarization
US9330084B1 (en) 2014-12-10 2016-05-03 International Business Machines Corporation Automatically generating question-answer pairs during content ingestion by a question answering computing system
US9367745B2 (en) 2012-04-24 2016-06-14 Liveclips Llc System for annotating media content for automatic content understanding
WO2017074562A1 (en) * 2015-10-28 2017-05-04 Intel Corporation Automatic video summarization
US9659597B2 (en) 2012-04-24 2017-05-23 Liveclips Llc Annotating media content for automatic content understanding
US9665824B2 (en) 2008-12-22 2017-05-30 The Trustees Of Columbia University In The City Of New York Rapid image annotation via brain state decoding and visual pattern mining
EP3110165A4 (en) * 2014-04-11 2017-08-09 Samsung Electronics Co., Ltd. Broadcast receiving apparatus and method for summarized content service
RU2639668C2 (en) * 2015-10-30 2017-12-21 Сяоми Инк. Method and device for region identification
WO2019195835A1 (en) * 2018-04-06 2019-10-10 Deluxe One Llc Comparing frame data to generate a textless version of a multimedia production
WO2020154554A1 (en) 2019-01-25 2020-07-30 Gracenote, Inc. Methods and systems for scoreboard region detection
WO2020154555A1 (en) * 2019-01-25 2020-07-30 Gracenote, Inc. Methods and systems for scoreboard text region detection
US10997424B2 (en) 2019-01-25 2021-05-04 Gracenote, Inc. Methods and systems for sport data extraction
US11087161B2 (en) 2019-01-25 2021-08-10 Gracenote, Inc. Methods and systems for determining accuracy of sport-related information extracted from digital video frames
US11805283B2 (en) 2019-01-25 2023-10-31 Gracenote, Inc. Methods and systems for extracting sport-related information from digital video frames

Families Citing this family (147)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US6735253B1 (en) * 1997-05-16 2004-05-11 The Trustees Of Columbia University In The City Of New York Methods and architecture for indexing and editing compressed video over the world wide web
US7143434B1 (en) 1998-11-06 2006-11-28 Seungyup Paek Video description system and method
US6263503B1 (en) 1999-05-26 2001-07-17 Neal Margulis Method for effectively implementing a wireless television system
US8266657B2 (en) 2001-03-15 2012-09-11 Sling Media Inc. Method for effectively implementing a multi-room television system
WO2003051031A2 (en) 2001-12-06 2003-06-19 The Trustees Of Columbia University In The City Of New York Method and apparatus for planarization of a material by growing and removing a sacrificial film
US7120873B2 (en) * 2002-01-28 2006-10-10 Sharp Laboratories Of America, Inc. Summarization of sumo video content
JP3744464B2 (en) * 2002-05-20 2006-02-08 ソニー株式会社 Signal recording / reproducing apparatus and method, signal reproducing apparatus and method, program, and recording medium
US7577841B2 (en) * 2002-08-15 2009-08-18 Digimarc Corporation Watermark placement in watermarking of time varying media signals
US20040152055A1 (en) * 2003-01-30 2004-08-05 Gliessner Michael J.G. Video based language learning system
KR101058054B1 (en) * 2003-08-18 2011-08-19 코닌클리케 필립스 일렉트로닉스 엔.브이. Extract video
JP4233982B2 (en) * 2003-11-06 2009-03-04 パイオニア株式会社 Image processing apparatus, image processing method, image processing program, and information recording medium recording the same
US7882436B2 (en) * 2004-03-10 2011-02-01 Trevor Burke Technology Limited Distribution of video data
WO2005107244A1 (en) * 2004-04-29 2005-11-10 Koninklijke Philips Electronics N.V. Method and receiver for controlling broadcast data
US7769756B2 (en) 2004-06-07 2010-08-03 Sling Media, Inc. Selection and presentation of context-relevant supplemental content and advertising
US9998802B2 (en) 2004-06-07 2018-06-12 Sling Media LLC Systems and methods for creating variable length clips from a media stream
US8346605B2 (en) 2004-06-07 2013-01-01 Sling Media, Inc. Management of shared media content
US7917932B2 (en) 2005-06-07 2011-03-29 Sling Media, Inc. Personal video recorder functionality for placeshifting systems
US7975062B2 (en) 2004-06-07 2011-07-05 Sling Media, Inc. Capturing and sharing media content
US8099755B2 (en) 2004-06-07 2012-01-17 Sling Media Pvt. Ltd. Systems and methods for controlling the encoding of a media stream
CN101243688A (en) 2004-06-07 2008-08-13 斯灵媒体公司 Personal media broadcasting system
FR2872660B1 (en) * 2004-07-05 2006-12-22 Eastman Kodak Co SHOOTING APPARATUS AND METHOD FOR FORMATION OF ANNOTATED IMAGES
US7765218B2 (en) * 2004-09-30 2010-07-27 International Business Machines Corporation Determining a term score for an animated graphics file
WO2006072897A1 (en) * 2005-01-04 2006-07-13 Koninklijke Philips Electronics N.V. Method and device for detecting transparent regions
WO2006096612A2 (en) 2005-03-04 2006-09-14 The Trustees Of Columbia University In The City Of New York System and method for motion estimation and mode decision for low-complexity h.264 decoder
US8064516B2 (en) * 2005-06-02 2011-11-22 Broadcom Corporation Text recognition during video compression
EP1899814B1 (en) 2005-06-30 2017-05-03 Sling Media, Inc. Firmware update for consumer electronic device
US20070011012A1 (en) * 2005-07-11 2007-01-11 Steve Yurick Method, system, and apparatus for facilitating captioning of multi-media content
US8874477B2 (en) 2005-10-04 2014-10-28 Steven Mark Hoffberg Multifactorial optimization system and method
US7752543B2 (en) * 2006-02-17 2010-07-06 Microsoft Corporation Applying effects to a merged text path
JP2007228220A (en) * 2006-02-23 2007-09-06 Funai Electric Co Ltd Built-in hard diskdrive television receiver and television receiver
KR100764175B1 (en) * 2006-02-27 2007-10-08 삼성전자주식회사 Apparatus and Method for Detecting Key Caption in Moving Picture for Customized Service
RU2440606C2 (en) * 2006-03-03 2012-01-20 Конинклейке Филипс Электроникс Н.В. Method and apparatus for automatic generation of summary of plurality of images
US7734092B2 (en) * 2006-03-07 2010-06-08 Ancestry.Com Operations Inc. Multiple image input for optical character recognition processing systems and methods
US20070242926A1 (en) * 2006-04-13 2007-10-18 Huang Chen-Hsiu Video Program Processing Method and System
US8392183B2 (en) 2006-04-25 2013-03-05 Frank Elmo Weber Character-based automated media summarization
US7532767B2 (en) * 2006-05-31 2009-05-12 Xerox Corporation Removing ringing and blocking artifacts from JPEG compressed document images
US8004608B2 (en) * 2006-06-08 2011-08-23 Shenzhen Tcl New Technology Ltd Closed captioning data detection system and method
GB2441010A (en) * 2006-08-17 2008-02-20 Green Cathedral Plc Creating a subtitle database
US20080111822A1 (en) * 2006-09-22 2008-05-15 Yahoo, Inc.! Method and system for presenting video
US7966552B2 (en) 2006-10-16 2011-06-21 Sony Corporation Trial selection of STB remote control codes
US7689613B2 (en) * 2006-10-23 2010-03-30 Sony Corporation OCR input to search engine
US8077263B2 (en) 2006-10-23 2011-12-13 Sony Corporation Decoding multiple remote control code sets
US7814524B2 (en) 2007-02-14 2010-10-12 Sony Corporation Capture of configuration and service provider data via OCR
US7991271B2 (en) 2007-02-14 2011-08-02 Sony Corporation Transfer of metadata using video frames
KR100836197B1 (en) * 2006-12-14 2008-06-09 삼성전자주식회사 Apparatus for detecting caption in moving picture and method of operating the apparatus
US8763038B2 (en) 2009-01-26 2014-06-24 Sony Corporation Capture of stylized TV table data via OCR
DE602007007369D1 (en) 2007-03-16 2010-08-05 Thomson Licensing SYSTEM AND METHOD FOR COMBINING TEXT WITH THREE-DIMENSIONAL CONTENTS
US8438589B2 (en) 2007-03-28 2013-05-07 Sony Corporation Obtaining metadata program information during channel changes
JP4945739B2 (en) * 2007-03-30 2012-06-06 日本電産サンキョー株式会社 Character string recognition method and character string recognition apparatus
US8238719B2 (en) * 2007-05-08 2012-08-07 Cyberlink Corp. Method for processing a sports video and apparatus thereof
US8094202B2 (en) * 2007-05-17 2012-01-10 Canon Kabushiki Kaisha Moving image capture apparatus and moving image capture method
US9351048B1 (en) 2007-08-29 2016-05-24 The Directv Group, Inc. Method and system for assigning a channel to data in a data stream
CN101382934B (en) * 2007-09-06 2010-08-18 华为技术有限公司 Search method for multimedia model, apparatus and system
US8477793B2 (en) 2007-09-26 2013-07-02 Sling Media, Inc. Media streaming device with gateway functionality
US8350971B2 (en) 2007-10-23 2013-01-08 Sling Media, Inc. Systems and methods for controlling media devices
US8060609B2 (en) * 2008-01-04 2011-11-15 Sling Media Inc. Systems and methods for determining attributes of media items accessed via a personal media broadcaster
WO2009126785A2 (en) 2008-04-10 2009-10-15 The Trustees Of Columbia University In The City Of New York Systems and methods for image archaeology
US8079054B1 (en) * 2008-04-14 2011-12-13 Adobe Systems Incorporated Location for secondary content based on data differential
US9326004B2 (en) 2008-06-03 2016-04-26 Broadcom Corporation Reduced memory mode video decode
WO2009155281A1 (en) 2008-06-17 2009-12-23 The Trustees Of Columbia University In The City Of New York System and method for dynamically and interactively searching media data
US8667279B2 (en) 2008-07-01 2014-03-04 Sling Media, Inc. Systems and methods for securely place shifting media content
US8463053B1 (en) 2008-08-08 2013-06-11 The Research Foundation Of State University Of New York Enhanced max margin learning on multimodal data mining in a multimedia database
US8381310B2 (en) 2009-08-13 2013-02-19 Sling Media Pvt. Ltd. Systems, methods, and program applications for selectively restricting the placeshifting of copy protected digital media content
US8320674B2 (en) 2008-09-03 2012-11-27 Sony Corporation Text localization for image and video OCR
US8667163B2 (en) * 2008-09-08 2014-03-04 Sling Media Inc. Systems and methods for projecting images from a computer system
US8396004B2 (en) * 2008-11-10 2013-03-12 At&T Intellectual Property Ii, L.P. Video share model-based video fixing
US8035656B2 (en) 2008-11-17 2011-10-11 Sony Corporation TV screen text capture
US9191610B2 (en) 2008-11-26 2015-11-17 Sling Media Pvt Ltd. Systems and methods for creating logical media streams for media storage and playback
US20100188457A1 (en) * 2009-01-05 2010-07-29 Madigan Connor F Method and apparatus for controlling the temperature of an electrically-heated discharge nozzle
US8438602B2 (en) 2009-01-26 2013-05-07 Sling Media Inc. Systems and methods for linking media content
US8425325B2 (en) * 2009-02-06 2013-04-23 Apple Inc. Automatically generating a book describing a user's videogame performance
JP5421627B2 (en) * 2009-03-19 2014-02-19 キヤノン株式会社 Video data display apparatus and method
US8559720B2 (en) * 2009-03-30 2013-10-15 Thomson Licensing S.A. Using a video processing and text extraction method to identify video segments of interest
US8171148B2 (en) * 2009-04-17 2012-05-01 Sling Media, Inc. Systems and methods for establishing connections between devices communicating over a network
US8406431B2 (en) 2009-07-23 2013-03-26 Sling Media Pvt. Ltd. Adaptive gain control for digital audio samples in a media stream
US9479737B2 (en) * 2009-08-06 2016-10-25 Echostar Technologies L.L.C. Systems and methods for event programming via a remote media player
US20110032986A1 (en) * 2009-08-07 2011-02-10 Sling Media Pvt Ltd Systems and methods for automatically controlling the resolution of streaming video content
US20110035765A1 (en) * 2009-08-10 2011-02-10 Sling Media Pvt Ltd Systems and methods for providing programming content
US8532472B2 (en) * 2009-08-10 2013-09-10 Sling Media Pvt Ltd Methods and apparatus for fast seeking within a media stream buffer
US8799408B2 (en) * 2009-08-10 2014-08-05 Sling Media Pvt Ltd Localization systems and methods
US9525838B2 (en) 2009-08-10 2016-12-20 Sling Media Pvt. Ltd. Systems and methods for virtual remote control of streamed media
US8966101B2 (en) 2009-08-10 2015-02-24 Sling Media Pvt Ltd Systems and methods for updating firmware over a network
US9565479B2 (en) 2009-08-10 2017-02-07 Sling Media Pvt Ltd. Methods and apparatus for seeking within a media stream using scene detection
US20110035466A1 (en) * 2009-08-10 2011-02-10 Sling Media Pvt Ltd Home media aggregator system and method
KR20110021195A (en) * 2009-08-25 2011-03-04 삼성전자주식회사 Method and apparatus for detecting an important information from a moving picture
US9160974B2 (en) * 2009-08-26 2015-10-13 Sling Media, Inc. Systems and methods for transcoding and place shifting media content
US8314893B2 (en) 2009-08-28 2012-11-20 Sling Media Pvt. Ltd. Remote control and method for automatically adjusting the volume output of an audio device
US8922718B2 (en) * 2009-10-21 2014-12-30 Disney Enterprises, Inc. Key generation through spatial detection of dynamic objects
US8327407B2 (en) 2009-10-27 2012-12-04 Sling Media, Inc. Determination of receiving live versus time-shifted media content at a communication device
US20110113354A1 (en) * 2009-11-12 2011-05-12 Sling Media Pvt Ltd Always-on-top media player launched from a web browser
US9015225B2 (en) 2009-11-16 2015-04-21 Echostar Technologies L.L.C. Systems and methods for delivering messages over a network
US8799485B2 (en) * 2009-12-18 2014-08-05 Sling Media, Inc. Methods and apparatus for establishing network connections using an inter-mediating device
US8626879B2 (en) 2009-12-22 2014-01-07 Sling Media, Inc. Systems and methods for establishing network connections using local mediation services
US9178923B2 (en) * 2009-12-23 2015-11-03 Echostar Technologies L.L.C. Systems and methods for remotely controlling a media server via a network
US9275054B2 (en) 2009-12-28 2016-03-01 Sling Media, Inc. Systems and methods for searching media content
US8989491B2 (en) * 2009-12-31 2015-03-24 Tata Consultancy Services Limited Method and system for preprocessing the region of video containing text
US8856349B2 (en) 2010-02-05 2014-10-07 Sling Media Inc. Connection priority services for data communication between two devices
US20120008693A1 (en) * 2010-07-08 2012-01-12 Echostar Technologies L.L.C. Substituting Embedded Text for Video Text Images
JP2012038239A (en) * 2010-08-11 2012-02-23 Sony Corp Information processing equipment, information processing method and program
US20120206567A1 (en) * 2010-09-13 2012-08-16 Trident Microsystems (Far East) Ltd. Subtitle detection system and method to television video
US8023697B1 (en) 2011-03-29 2011-09-20 Kaspersky Lab Zao System and method for identifying spam in rasterized images
WO2012143596A1 (en) * 2011-04-18 2012-10-26 Supponor Oy Detection of graphics added to a video signal
US9373039B2 (en) * 2011-04-18 2016-06-21 Supponor Oy Detection of graphics added to a video signal
TWI431516B (en) * 2011-06-21 2014-03-21 Quanta Comp Inc Method and electronic device for tactile feedback
KR101975247B1 (en) * 2011-09-14 2019-08-23 삼성전자주식회사 Image processing apparatus and image processing method thereof
US9704111B1 (en) 2011-09-27 2017-07-11 3Play Media, Inc. Electronic transcription job market
US8918311B1 (en) * 2012-03-21 2014-12-23 3Play Media, Inc. Intelligent caption systems and methods
US20130268621A1 (en) * 2012-04-08 2013-10-10 Broadcom Corporation Transmission of video utilizing static content information from video source
US9154749B2 (en) * 2012-04-08 2015-10-06 Broadcom Corporation Power saving techniques for wireless delivery of video
US8861858B2 (en) 2012-06-01 2014-10-14 Blackberry Limited Methods and devices for providing companion services to video
US8819759B2 (en) * 2012-06-27 2014-08-26 Google Technology Holdings LLC Determining the location of a point of interest in a media stream that includes caption data
US9087030B2 (en) * 2012-07-16 2015-07-21 International Business Machines Corporation Handling excessive input characters in a field
IL223381B (en) * 2012-12-02 2018-01-31 Berale Of Teldan Group Ltd Automatic summarising of media content
KR102123062B1 (en) 2013-08-06 2020-06-15 삼성전자주식회사 Method of aquiring information about contents, image display apparatus using thereof and server system of providing information about contents
US9693030B2 (en) 2013-09-09 2017-06-27 Arris Enterprises Llc Generating alerts based upon detector outputs
CA2924065C (en) 2013-09-13 2018-05-15 Arris Enterprises, Inc. Content based video content segmentation
US9456170B1 (en) 2013-10-08 2016-09-27 3Play Media, Inc. Automated caption positioning systems and methods
US10319035B2 (en) 2013-10-11 2019-06-11 Ccc Information Services Image capturing and automatic labeling system
CN103761345A (en) * 2014-02-27 2014-04-30 苏州千视通信科技有限公司 Video retrieval method based on OCR character recognition technology
JP6464616B2 (en) * 2014-08-27 2019-02-06 富士通株式会社 Information processing program, method, and apparatus
JP6394184B2 (en) * 2014-08-27 2018-09-26 富士通株式会社 Judgment program, method, and apparatus
US9582727B2 (en) * 2015-01-16 2017-02-28 Sony Corporation Text recognition system with feature recognition and method of operation thereof
US11076198B2 (en) * 2015-05-28 2021-07-27 Idomoo Ltd. System and method to generate an interactive video on the fly
US9471990B1 (en) * 2015-10-20 2016-10-18 Interra Systems, Inc. Systems and methods for detection of burnt-in text in a video
US9805269B2 (en) * 2015-11-20 2017-10-31 Adobe Systems Incorporated Techniques for enhancing content memorability of user generated video content
WO2017088478A1 (en) * 2015-11-24 2017-06-01 乐视控股(北京)有限公司 Number separating method and device
CN105868683A (en) * 2015-11-24 2016-08-17 乐视致新电子科技(天津)有限公司 Channel logo identification method and apparatus
US10963504B2 (en) * 2016-02-12 2021-03-30 Sri International Zero-shot event detection using semantic embedding
US10575036B2 (en) 2016-03-02 2020-02-25 Google Llc Providing an indication of highlights in a video content item
CN106162223B (en) * 2016-05-27 2020-06-05 北京奇虎科技有限公司 News video segmentation method and device
JP6910875B2 (en) * 2016-07-22 2021-07-28 ジョセフ プローグマンJoseph Ploegman Systems and methods for displaying baseball batter counts
US9972360B2 (en) * 2016-08-30 2018-05-15 Oath Inc. Computerized system and method for automatically generating high-quality digital content thumbnails from digital video
GB2558868A (en) * 2016-09-29 2018-07-25 British Broadcasting Corp Video search system & method
US11594028B2 (en) 2018-05-18 2023-02-28 Stats Llc Video processing for enabling sports highlights generation
CN109583443B (en) * 2018-11-15 2022-10-18 四川长虹电器股份有限公司 Video content judgment method based on character recognition
CN111368838A (en) * 2018-12-26 2020-07-03 珠海金山网络游戏科技有限公司 Method and device for identifying reported screenshot
US10701434B1 (en) * 2019-01-21 2020-06-30 Adobe Inc. Extracting session information from video content to facilitate seeking
CN110083741B (en) * 2019-04-11 2022-10-28 中国科学技术大学 Character-oriented video abstract extraction method based on text and image combined modeling
CN110222168B (en) * 2019-05-20 2023-08-18 平安科技(深圳)有限公司 Data processing method and related device
CN111988652B (en) * 2019-05-23 2022-06-03 北京地平线机器人技术研发有限公司 Method and device for extracting lip language training data
US11087162B2 (en) * 2019-08-01 2021-08-10 Nvidia Corporation Determining relative regions of interest in images using object detection
CN110929094B (en) * 2019-11-20 2023-05-16 北京香侬慧语科技有限责任公司 Video title processing method and device
EP3923189A1 (en) * 2020-06-08 2021-12-15 Interdigital Ce Patent Holdings, SAS Generation of metadata from graphical inlays inserted in video frames
CN111491180B (en) * 2020-06-24 2021-07-09 腾讯科技(深圳)有限公司 Method and device for determining key frame
US11554324B2 (en) * 2020-06-25 2023-01-17 Sony Interactive Entertainment LLC Selection of video template based on computer simulation metadata
US11735186B2 (en) 2021-09-07 2023-08-22 3Play Media, Inc. Hybrid live captioning systems and methods
CN114615520B (en) * 2022-03-08 2024-01-02 北京达佳互联信息技术有限公司 Subtitle positioning method, subtitle positioning device, computer equipment and medium

Family Cites Families (258)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
GB2141607A (en) * 1983-06-15 1984-12-19 Philips Electronic Associated Video display system with index pages
JPS60204086A (en) 1984-03-28 1985-10-15 Fuji Electric Co Ltd Object discriminating device
US5579444A (en) 1987-08-28 1996-11-26 Axiom Bildverarbeitungssysteme Gmbh Adaptive vision-based controller
US5664177A (en) 1988-04-13 1997-09-02 Digital Equipment Corporation Data processing system having a data structure with a single, simple primitive
CA2001390C (en) 1988-12-19 1997-12-30 Ming-Chien Shan View composition in a data-base management system
US5144685A (en) 1989-03-31 1992-09-01 Honeywell Inc. Landmark recognition for autonomous mobile robots
US5682425A (en) * 1990-04-23 1997-10-28 Canon Kabushiki Kaisha Information signal transmission system
FR2661585B1 (en) * 1990-04-25 1992-08-07 Telediffusion Fse METHOD AND DEVICE FOR SCRAMBLING-SCRATCHING DIGITAL IMAGE DATA.
JPH04207878A (en) * 1990-11-30 1992-07-29 Toshiba Corp Moving image management device
US5191645A (en) * 1991-02-28 1993-03-02 Sony Corporation Of America Digital signal processing system employing icon displays
JP3177746B2 (en) * 1991-03-20 2001-06-18 株式会社日立製作所 Data processing system and method
US5212742A (en) 1991-05-24 1993-05-18 Apple Computer, Inc. Method and apparatus for encoding/decoding image data
EP0528084B1 (en) 1991-08-15 1999-05-19 International Business Machines Corporation System and method for processing data representing stored images
US5546572A (en) 1991-08-28 1996-08-13 Hitachi, Ltd. Method for retrieving database of image information
US5701510A (en) 1991-11-14 1997-12-23 International Business Machines Corporation Method and system for efficient designation and retrieval of particular segments within a multimedia presentation utilizing a data processing system
US7006881B1 (en) * 1991-12-23 2006-02-28 Steven Hoffberg Media recording device with remote graphic user interface
JPH0756652B2 (en) 1992-03-24 1995-06-14 インターナショナル・ビジネス・マシーンズ・コーポレイション Search for video frame sequence
US5623690A (en) * 1992-06-03 1997-04-22 Digital Equipment Corporation Audio/video storage and retrieval for multimedia workstations by interleaving audio and video data in data file
US5262856A (en) 1992-06-04 1993-11-16 Massachusetts Institute Of Technology Video image compositing techniques
GB9215102D0 (en) 1992-07-16 1992-08-26 Philips Electronics Uk Ltd Tracking moving objects
US5721788A (en) 1992-07-31 1998-02-24 Corbis Corporation Method and system for digital image signatures
GB2270436A (en) 1992-09-05 1994-03-09 Ibm Target tracking system
JP2903904B2 (en) * 1992-10-09 1999-06-14 松下電器産業株式会社 Image retrieval device
US5579471A (en) 1992-11-09 1996-11-26 International Business Machines Corporation Image query system and method
US5615112A (en) * 1993-01-29 1997-03-25 Arizona Board Of Regents Synthesized object-oriented entity-relationship (SOOER) model for coupled knowledge-base/database of image retrieval expert system (IRES)
US5630121A (en) * 1993-02-02 1997-05-13 International Business Machines Corporation Archiving and retrieving multimedia objects using structured indexes
US5408274A (en) * 1993-03-11 1995-04-18 The Regents Of The University Of California Method and apparatus for compositing compressed video data
US5555354A (en) 1993-03-23 1996-09-10 Silicon Graphics Inc. Method and apparatus for navigation within three-dimensional information landscape
US5391199A (en) 1993-07-20 1995-02-21 Biosense, Inc. Apparatus and method for treating cardiac arrhythmias
US5619709A (en) 1993-09-20 1997-04-08 Hnc, Inc. System and method of context vector generation and retrieval
JP3303543B2 (en) * 1993-09-27 2002-07-22 インターナショナル・ビジネス・マシーンズ・コーポレーション How to organize and play multimedia segments, and how to organize and play two or more multimedia stories as hyperstory
US6614914B1 (en) * 1995-05-08 2003-09-02 Digimarc Corporation Watermark embedder and reader
US6122403A (en) 1995-07-27 2000-09-19 Digimarc Corporation Computer system linked by using information in data objects
US5600831A (en) 1994-02-28 1997-02-04 Lucent Technologies Inc. Apparatus and methods for retrieving information by modifying query plan based on description of information sources
US5555378A (en) 1994-03-02 1996-09-10 Bell Communications Research, Inc. Scheduling transmission multimedia information in broadband networks using a token passing scheme
US5668897A (en) 1994-03-15 1997-09-16 Stolfo; Salvatore J. Method and apparatus for imaging, image processing and data compression merge/purge techniques for document image databases
US5606655A (en) * 1994-03-31 1997-02-25 Siemens Corporate Research, Inc. Method for representing contents of a single video shot using frames
US5465353A (en) 1994-04-01 1995-11-07 Ricoh Company, Ltd. Image matching and retrieval by multi-access redundant hashing
TW259725B (en) * 1994-04-11 1995-10-11 Mitsubishi Heavy Ind Ltd
US5488664A (en) * 1994-04-22 1996-01-30 Yeda Research And Development Co., Ltd. Method and apparatus for protecting visual information with printed cryptographic watermarks
US5493677A (en) * 1994-06-08 1996-02-20 Systems Research & Applications Corporation Generation, archiving, and retrieval of digital images with evoked suggestion-set captions and natural language interface
US5613032A (en) * 1994-09-02 1997-03-18 Bell Communications Research, Inc. System and method for recording, playing back and searching multimedia events wherein video, audio and text can be searched and retrieved
WO1996008095A1 (en) 1994-09-08 1996-03-14 Virtex Communications, Inc. Method and apparatus for electronic distribution of digital multi-media information
US5642477A (en) 1994-09-22 1997-06-24 International Business Machines Corporation Method and apparatus for selectably retrieving and outputting digitally stored multimedia presentations with real-time non-interrupting, dynamically selectable introduction of output processing
US5802361A (en) 1994-09-30 1998-09-01 Apple Computer, Inc. Method and system for searching graphic images and videos
US5566089A (en) 1994-10-26 1996-10-15 General Instrument Corporation Of Delaware Syntax parser for a video decompression processor
DE69535098T2 (en) 1994-11-15 2007-04-19 Canon K.K. Method and apparatus for searching images in a database
WO1996017313A1 (en) 1994-11-18 1996-06-06 Oracle Corporation Method and apparatus for indexing multimedia information streams
WO1996017306A2 (en) 1994-11-21 1996-06-06 Oracle Corporation Media server
US5805733A (en) 1994-12-12 1998-09-08 Apple Computer, Inc. Method and system for detecting scenes and summarizing video sequences
US5646997A (en) * 1994-12-14 1997-07-08 Barton; James M. Method and apparatus for embedding authentication information within digital data
US5530759A (en) 1995-02-01 1996-06-25 International Business Machines Corporation Color correct digital watermarking of images
US5821945A (en) 1995-02-03 1998-10-13 The Trustees Of Princeton University Method and apparatus for video browsing based on content and structure
US5794242A (en) 1995-02-07 1998-08-11 Digital Equipment Corporation Temporally and spatially organized database
US5943422A (en) 1996-08-12 1999-08-24 Intertrust Technologies Corp. Steganographic techniques for securely delivering electronic digital rights management control information over insecure communication channels
JPH08262965A (en) 1995-03-20 1996-10-11 Mitsubishi Electric Corp Closed caption decoder with pause function for language learning
US5684715A (en) 1995-06-07 1997-11-04 Canon Information Systems, Inc. Interactive video system with dynamic video object descriptors
US5721815A (en) * 1995-06-07 1998-02-24 International Business Machines Corporation Media-on-demand communication system and method employing direct access storage device
JPH0916457A (en) * 1995-06-28 1997-01-17 Fujitsu Ltd Multimedia data retrieval system
US5758076A (en) * 1995-07-19 1998-05-26 International Business Machines Corporation Multimedia server system having rate adjustable data retrieval based on buffer capacity
US5822524A (en) 1995-07-21 1998-10-13 Infovalue Computing, Inc. System for just-in-time retrieval of multimedia files over computer networks by transmitting data packets at transmission rate determined by frame size
EP0757490A3 (en) * 1995-08-02 1999-01-13 Matsushita Electric Industrial Co., Ltd. Video coding device and video transmission system using the same, quantization control method and average throughput calculation method used therein
US5734893A (en) * 1995-09-28 1998-03-31 Ibm Corporation Progressive content-based retrieval of image and video with adaptive and iterative refinement
US5969755A (en) 1996-02-05 1999-10-19 Texas Instruments Incorporated Motion based event detection system and method
US5664018A (en) 1996-03-12 1997-09-02 Leighton; Frank Thomson Watermarking process resilient to collusion attacks
US5949885A (en) 1996-03-12 1999-09-07 Leighton; F. Thomson Method for protecting content using watermarking
US5987459A (en) 1996-03-15 1999-11-16 Regents Of The University Of Minnesota Image and document management system for content-based retrieval
US5884298A (en) * 1996-03-29 1999-03-16 Cygnet Storage Solutions, Inc. Method for accessing and updating a library of optical discs
US5893095A (en) * 1996-03-29 1999-04-06 Virage, Inc. Similarity engine for content-based retrieval of images
US5767922A (en) 1996-04-05 1998-06-16 Cornell Research Foundation, Inc. Apparatus and process for detecting scene breaks in a sequence of video frames
US5852435A (en) 1996-04-12 1998-12-22 Avid Technology, Inc. Digital multimedia editing and data management system
US5696964A (en) 1996-04-16 1997-12-09 Nec Research Institute, Inc. Multimedia database retrieval system which maintains a posterior probability distribution that each item in the database is a target of a search
US5870754A (en) * 1996-04-25 1999-02-09 Philips Electronics North America Corporation Video retrieval of MPEG compressed sequences using DC and motion signatures
JP3423835B2 (en) * 1996-05-01 2003-07-07 沖電気工業株式会社 Compression encoding device with scramble and decompression reproduction device thereof
MY119560A (en) * 1996-05-27 2005-06-30 Nippon Telegraph & Telephone Scheme for detecting captions in coded video data without decoding coded video data
US6078664A (en) 1996-12-20 2000-06-20 Moskowitz; Scott A. Z-transform implementation of digital watermarks
US6154755A (en) 1996-07-31 2000-11-28 Eastman Kodak Company Index imaging system
US6031914A (en) * 1996-08-30 2000-02-29 Regents Of The University Of Minnesota Method and apparatus for embedding data, including watermarks, in human perceptible images
US6282299B1 (en) 1996-08-30 2001-08-28 Regents Of The University Of Minnesota Method and apparatus for video watermarking using perceptual masks
US5848155A (en) 1996-09-04 1998-12-08 Nec Research Institute, Inc. Spread spectrum watermark for embedded signalling
US5809139A (en) 1996-09-13 1998-09-15 Vivo Software, Inc. Watermarking method and apparatus for compressed digital video
US5873080A (en) * 1996-09-20 1999-02-16 International Business Machines Corporation Using multiple search engines to search multimedia data
US5734752A (en) * 1996-09-24 1998-03-31 Xerox Corporation Digital watermarking using stochastic screen patterns
US6075875A (en) 1996-09-30 2000-06-13 Microsoft Corporation Segmentation of image features using hierarchical analysis of multi-valued image data and weighted averaging of segmentation results
US5852823A (en) 1996-10-16 1998-12-22 Microsoft Image classification and retrieval system using a query-by-example paradigm
SE515535C2 (en) * 1996-10-25 2001-08-27 Ericsson Telefon Ab L M A transcoder
US5825892A (en) 1996-10-28 1998-10-20 International Business Machines Corporation Protecting images with an image watermark
US5915027A (en) 1996-11-05 1999-06-22 Nec Research Institute Digital watermarking
US6172675B1 (en) * 1996-12-05 2001-01-09 Interval Research Corporation Indirect manipulation of data using temporally related data, with particular application to manipulation of audio or audiovisual data
US5953506A (en) 1996-12-17 1999-09-14 Adaptive Media Technologies Method and apparatus that provides a scalable media delivery system
JP3106985B2 (en) 1996-12-25 2000-11-06 日本電気株式会社 Electronic watermark insertion device and detection device
US6058205A (en) * 1997-01-09 2000-05-02 International Business Machines Corporation System and method for partitioning the feature space of a classifier in a pattern classification system
US5963203A (en) 1997-07-03 1999-10-05 Obvious Technology, Inc. Interactive video icon with designated viewing position
US5790703A (en) 1997-01-21 1998-08-04 Xerox Corporation Digital watermarking using conjugate halftone screens
US5901245A (en) 1997-01-23 1999-05-04 Eastman Kodak Company Method and system for detection and characterization of open space in digital images
US6157746A (en) 1997-02-12 2000-12-05 Sarnoff Corporation Apparatus and method for encoding wavelet trees generated by a wavelet-based coding method
AUPO525497A0 (en) 1997-02-21 1997-03-20 Mills, Dudley John Network-based classified information systems
US5930783A (en) 1997-02-21 1999-07-27 Nec Usa, Inc. Semantic and cognition based image retrieval
JP3736706B2 (en) 1997-04-06 2006-01-18 ソニー株式会社 Image display apparatus and method
WO1998046005A2 (en) 1997-04-07 1998-10-15 At & T Corp. System and method for processing object-based audiovisual information
US5937422A (en) 1997-04-15 1999-08-10 The United States Of America As Represented By The National Security Agency Automatically generating a topic description for text and searching and sorting text by topic using the same
JP3485438B2 (en) 1997-04-16 2004-01-13 シャープ株式会社 Electronic calculator with graph display function
US6741655B1 (en) 1997-05-05 2004-05-25 The Trustees Of Columbia University In The City Of New York Algorithms and system for object-oriented content-based video search
US6208746B1 (en) * 1997-05-09 2001-03-27 Gte Service Corporation Biometric watermarks
US6735253B1 (en) * 1997-05-16 2004-05-11 The Trustees Of Columbia University In The City Of New York Methods and architecture for indexing and editing compressed video over the world wide web
US6125229A (en) 1997-06-02 2000-09-26 Philips Electronics North America Corporation Visual indexing system
US6131162A (en) 1997-06-05 2000-10-10 Hitachi Ltd. Digital data authentication method
US5960081A (en) 1997-06-05 1999-09-28 Cray Research, Inc. Embedding a digital signature in a video sequence
US6222932B1 (en) * 1997-06-27 2001-04-24 International Business Machines Corporation Automatic adjustment of image watermark strength based on computed image texture
US5983218A (en) 1997-06-30 1999-11-09 Xerox Corporation Multimedia database for use over networks
US6195458B1 (en) * 1997-07-29 2001-02-27 Eastman Kodak Company Method for content-based temporal segmentation of video
US6360234B2 (en) * 1997-08-14 2002-03-19 Virage, Inc. Video cataloger system with synchronized encoders
US6208735B1 (en) * 1997-09-10 2001-03-27 Nec Research Institute, Inc. Secure spread spectrum watermarking for multimedia data
US6886013B1 (en) * 1997-09-11 2005-04-26 International Business Machines Corporation HTTP caching proxy to filter and control display of data in a web browser
US6108434A (en) 1997-09-12 2000-08-22 Signafy, Inc. Counteracting geometric distortions for DCT based watermarking
US6249318B1 (en) 1997-09-12 2001-06-19 8×8, Inc. Video coding/decoding arrangement and method therefor
US6807231B1 (en) 1997-09-12 2004-10-19 8×8, Inc. Multi-hypothesis motion-compensated video image predictor
US5995978A (en) 1997-09-24 1999-11-30 Ricoh Company, Ltd. Navigation system for document image database
US6070167A (en) * 1997-09-29 2000-05-30 Sharp Laboratories Of America, Inc. Hierarchical method and system for object-based audiovisual descriptive tagging of images for information retrieval, editing, and manipulation
US6070228A (en) * 1997-09-30 2000-05-30 International Business Machines Corp. Multimedia data storage system and method for operating a media server as a cache device and controlling a volume of data in the media server based on user-defined parameters
WO1999022318A1 (en) * 1997-10-27 1999-05-06 Massachusetts Institute Of Technology Image search and retrieval system
US7154560B1 (en) 1997-10-27 2006-12-26 Shih-Fu Chang Watermarking of digital image data
JPH11196386A (en) 1997-10-30 1999-07-21 Toshiba Corp Computer system and closed caption display method
US6072542A (en) 1997-11-25 2000-06-06 Fuji Xerox Co., Ltd. Automatic video segmentation using hidden markov model
FR2772538B1 (en) * 1997-12-17 2000-02-11 Telediffusion Fse METHOD AND SYSTEM FOR MEASURING THE QUALITY OF DIGITAL TELEVISION SIGNALS
US6100930A (en) 1997-12-19 2000-08-08 Thomson Consumer Electronics, Inc. Process and apparatus for performing wipes on compressed MPEG video bitstreams
US5956026A (en) 1997-12-19 1999-09-21 Sharp Laboratories Of America, Inc. Method for hierarchical summarization and browsing of digital video
US6037984A (en) * 1997-12-24 2000-03-14 Sarnoff Corporation Method and apparatus for embedding a watermark into a digital image or image sequence
US6208745B1 (en) 1997-12-30 2001-03-27 Sarnoff Corporation Method and apparatus for imbedding a watermark into a bitstream representation of a digital image sequence
US6332030B1 (en) 1998-01-15 2001-12-18 The Regents Of The University Of California Method for embedding and extracting digital data in images and video
US6654931B1 (en) 1998-01-27 2003-11-25 At&T Corp. Systems and methods for playing, browsing and interacting with MPEG-4 coded audio-visual objects
JP3738939B2 (en) 1998-03-05 2006-01-25 Kddi株式会社 Moving image cut point detection device
US6628824B1 (en) 1998-03-20 2003-09-30 Ken Belanger Method and apparatus for image identification and comparison
US6064764A (en) * 1998-03-30 2000-05-16 Seiko Epson Corporation Fragile watermarks for detecting tampering in images
US6092072A (en) 1998-04-07 2000-07-18 Lucent Technologies, Inc. Programmed medium for clustering large databases
US6240424B1 (en) * 1998-04-22 2001-05-29 Nbc Usa, Inc. Method and system for similarity-based image classification
US6504941B2 (en) 1998-04-30 2003-01-07 Hewlett-Packard Company Method and apparatus for digital watermarking of images
US6487301B1 (en) 1998-04-30 2002-11-26 Mediasec Technologies Llc Digital authentication with digital and analog documents
US7444308B2 (en) * 2001-06-15 2008-10-28 Health Discovery Corporation Data mining platform for bioinformatics and other knowledge discovery
CA2269651A1 (en) 1998-05-12 1999-11-12 Lucent Technologies, Inc. Transform domain image watermarking method and system
US6581058B1 (en) 1998-05-22 2003-06-17 Microsoft Corporation Scalable system for clustering of large databases having mixed data attributes
JP3615657B2 (en) 1998-05-27 2005-02-02 株式会社日立製作所 Video search method and apparatus, and recording medium
US6081278A (en) 1998-06-11 2000-06-27 Chen; Shenchang Eric Animation object having multiple resolution format
US6178416B1 (en) * 1998-06-15 2001-01-23 James U. Parker Method and apparatus for knowledgebase searching
US6285995B1 (en) 1998-06-22 2001-09-04 U.S. Philips Corporation Image retrieval system using a query image
US6476814B1 (en) 1998-06-25 2002-11-05 Wordgraph, Inc. Display structure for representation of complex systems
KR20010034920A (en) * 1998-06-26 2001-04-25 매클린토크 샤운 엘 Terminal for composing and presenting mpeg-4 video programs
US6041079A (en) * 1998-06-30 2000-03-21 Thomson Consumer Electronics, Inc, Field/frame conversion of DCT domain mixed field/frame mode macroblocks using 1-dimensional DCT/IDCT
CN1147079C (en) 1998-07-17 2004-04-21 皇家菲利浦电子有限公司 Device for demultiplexing coded data
US6226618B1 (en) 1998-08-13 2001-05-01 International Business Machines Corporation Electronic content delivery system
US6714909B1 (en) * 1998-08-13 2004-03-30 At&T Corp. System and method for automated multimedia content indexing and retrieval
US6346124B1 (en) * 1998-08-25 2002-02-12 University Of Florida Autonomous boundary detection system for echocardiographic images
US6708055B2 (en) * 1998-08-25 2004-03-16 University Of Florida Method for automated analysis of apical four-chamber images of the heart
JP3563270B2 (en) 1998-08-28 2004-09-08 株式会社日立製作所 Authentication method of digital watermark image
US6275599B1 (en) 1998-08-28 2001-08-14 International Business Machines Corporation Compressed image authentication and verification
US6970602B1 (en) 1998-10-06 2005-11-29 International Business Machines Corporation Method and apparatus for transcoding multimedia using content analysis
US6185329B1 (en) * 1998-10-13 2001-02-06 Hewlett-Packard Company Automatic caption text detection and processing for digital images
US6549911B2 (en) * 1998-11-02 2003-04-15 Survivors Of The Shoah Visual History Foundation Method and apparatus for cataloguing multimedia data
US6385602B1 (en) * 1998-11-03 2002-05-07 E-Centives, Inc. Presentation of search results using dynamic categorization
WO2000028440A1 (en) 1998-11-06 2000-05-18 The Trustees Of Columbia University In The City Ofnew York Systems and methods for interoperable multimedia content descriptions
US7143434B1 (en) 1998-11-06 2006-11-28 Seungyup Paek Video description system and method
US7254285B1 (en) 1998-11-06 2007-08-07 Seungup Paek Image description system and method
US6252975B1 (en) 1998-12-17 2001-06-26 Xerox Corporation Method and system for real time feature based motion analysis for key frame selection from a video
US6678389B1 (en) * 1998-12-29 2004-01-13 Kent Ridge Digital Labs Method and apparatus for embedding digital information in digital multimedia data
US6327390B1 (en) 1999-01-14 2001-12-04 Mitsubishi Electric Research Laboratories, Inc. Methods of scene fade detection for indexing of video sequences
US6532541B1 (en) * 1999-01-22 2003-03-11 The Trustees Of Columbia University In The City Of New York Method and apparatus for image authentication
US6643387B1 (en) 1999-01-28 2003-11-04 Sarnoff Corporation Apparatus and method for context-based indexing and retrieval of image sequences
US6366701B1 (en) * 1999-01-28 2002-04-02 Sarnoff Corporation Apparatus and method for describing the motion parameters of an object in an image sequence
US6223183B1 (en) * 1999-01-29 2001-04-24 International Business Machines Corporation System and method for describing views in space, time, frequency, and resolution
US6941325B1 (en) 1999-02-01 2005-09-06 The Trustees Of Columbia University Multimedia archive description scheme
US7185049B1 (en) * 1999-02-01 2007-02-27 At&T Corp. Multimedia integration description scheme, method and system for MPEG-7
US6236395B1 (en) * 1999-02-01 2001-05-22 Sharp Laboratories Of America, Inc. Audiovisual information management system
US6556695B1 (en) * 1999-02-05 2003-04-29 Mayo Foundation For Medical Education And Research Method for producing high resolution real-time images, of structure and function during medical procedures
US6404925B1 (en) 1999-03-11 2002-06-11 Fuji Xerox Co., Ltd. Methods and apparatuses for segmenting an audio-visual recording using image similarity searching and audio speaker recognition
US6556958B1 (en) * 1999-04-23 2003-04-29 Microsoft Corporation Fast clustering with sparse data
US6847980B1 (en) * 1999-07-03 2005-01-25 Ana B. Benitez Fundamental entity-relationship models for the generic audio visual data signal description
US6393394B1 (en) * 1999-07-19 2002-05-21 Qualcomm Incorporated Method and apparatus for interleaving line spectral information quantization methods in a speech coder
US6594629B1 (en) 1999-08-06 2003-07-15 International Business Machines Corporation Methods and apparatus for audio-visual speech detection and recognition
US6546135B1 (en) * 1999-08-30 2003-04-08 Mitsubishi Electric Research Laboratories, Inc Method for representing and comparing multimedia content
US6339450B1 (en) * 1999-09-21 2002-01-15 At&T Corp Error resilient transcoding for video over wireless channels
US7072398B2 (en) 2000-12-06 2006-07-04 Kai-Kuang Ma System and method for motion vector generation and analysis of digital video clips
US6725372B1 (en) * 1999-12-02 2004-04-20 Verizon Laboratories Inc. Digital watermarking
US6606393B1 (en) 1999-12-02 2003-08-12 Verizon Laboratories Inc. Message authentication code using image histograms
US6820111B1 (en) * 1999-12-07 2004-11-16 Microsoft Corporation Computer user interface architecture that saves a user's non-linear navigation history and intelligently maintains that history
DE19960372A1 (en) 1999-12-14 2001-06-21 Definiens Ag Process for processing data structures
US6282300B1 (en) 2000-01-21 2001-08-28 Signafy, Inc. Rotation, scale, and translation resilient public watermarking for images using a log-polar fourier transform
US6385329B1 (en) * 2000-02-14 2002-05-07 Digimarc Corporation Wavelet domain watermarks
AU2001238354A1 (en) * 2000-02-18 2001-08-27 The University Of Maryland Methods for the electronic annotation, retrieval, and use of electronic images
KR20010087552A (en) 2000-03-07 2001-09-21 구자홍 Dissolve/fade detection method for mpeg-compressed video using spatio-temporal distribution of the macro blocks
CN101035277A (en) 2000-03-13 2007-09-12 索尼公司 Method and apparatus for generating compact code-switching hints metadata
US6701309B1 (en) * 2000-04-21 2004-03-02 Lycos, Inc. Method and system for collecting related queries
US6567805B1 (en) 2000-05-15 2003-05-20 International Business Machines Corporation Interactive automated response system
US7177358B2 (en) 2000-06-27 2007-02-13 Mitsubishi Denki Kabushiki Kaisha Picture coding apparatus, and picture coding method
AU2001283004A1 (en) * 2000-07-24 2002-02-05 Vivcom, Inc. System and method for indexing, searching, identifying, and editing portions of electronic multimedia files
US20020157116A1 (en) 2000-07-28 2002-10-24 Koninklijke Philips Electronics N.V. Context and content based information processing for multimedia segmentation and indexing
US8538770B2 (en) * 2000-08-01 2013-09-17 Logical Images, Inc. System and method to aid diagnoses using cross-referenced knowledge and image databases
US6714683B1 (en) * 2000-08-24 2004-03-30 Digimarc Corporation Wavelet based feature modulation watermarks and related applications
WO2002025588A2 (en) * 2000-09-21 2002-03-28 Md Online Inc. Medical image processing systems
DE10048029A1 (en) 2000-09-26 2002-04-25 Philips Corp Intellectual Pty Procedure for calculating a transformation connecting two images
US6904408B1 (en) 2000-10-19 2005-06-07 Mccarthy John Bionet method, system and personalized web content manager responsive to browser viewers' psychological preferences, behavioral responses and physiological stress indicators
US7398275B2 (en) 2000-10-20 2008-07-08 Sony Corporation Efficient binary coding scheme for multimedia content descriptions
GB0029880D0 (en) 2000-12-07 2001-01-24 Sony Uk Ltd Video and audio information processing
US6879703B2 (en) 2001-01-10 2005-04-12 Trustees Of Columbia University Of The City Of New York Method and apparatus for watermarking images
US6915241B2 (en) * 2001-04-20 2005-07-05 Fraunhofer-Gesellschaft Zur Foerderung Der Angewandten Forschung E.V. Method for segmentation and identification of nonstationary time series
US6792434B2 (en) 2001-04-20 2004-09-14 Mitsubishi Electric Research Laboratories, Inc. Content-based visualization and user-modeling for interactive browsing and retrieval in multimedia databases
US20020169771A1 (en) 2001-05-09 2002-11-14 Melmon Kenneth L. System & method for facilitating knowledge management
ES2187353B1 (en) 2001-06-15 2004-08-16 Consultores Ub 57, S.L. DYNAMIC NAVIGATION INTERFACE.
US7145946B2 (en) 2001-07-27 2006-12-05 Sony Corporation MPEG video drift reduction
WO2003019416A1 (en) * 2001-08-28 2003-03-06 Koninklijke Philips Electronics N.V. Automatic question formulation from a user selection in multimedia content
WO2003051031A2 (en) 2001-12-06 2003-06-19 The Trustees Of Columbia University In The City Of New York Method and apparatus for planarization of a material by growing and removing a sacrificial film
US6700935B2 (en) * 2002-02-08 2004-03-02 Sony Electronics, Inc. Stream based bitrate transcoder for MPEG coded video
KR100959470B1 (en) 2002-03-22 2010-05-25 마이클 에프. 디어링 Scalable high performance 3d graphics
US7043474B2 (en) 2002-04-15 2006-05-09 International Business Machines Corporation System and method for measuring image similarity based on semantic meaning
JP2005525011A (en) 2002-04-26 2005-08-18 ザ トラスティーズ オブ コロンビア ユニヴァーシティ イン ザ シティ オブ ニューヨーク Method and system for optimal video transcoding based on utility function description
US20030229278A1 (en) 2002-06-06 2003-12-11 Usha Sinha Method and system for knowledge extraction from image data
JP3944016B2 (en) 2002-07-18 2007-07-11 パロマ工業株式会社 rice cooker
CA2493105A1 (en) 2002-07-19 2004-01-29 British Telecommunications Public Limited Company Method and system for classification of semantic content of audio/video data
JP3988599B2 (en) * 2002-09-20 2007-10-10 富士ゼロックス株式会社 Image processing method, falsification detection method, image processing apparatus, falsification detection apparatus, image processing program, falsification detection program, and image forming medium
US7103225B2 (en) 2002-11-07 2006-09-05 Honda Motor Co., Ltd. Clustering appearances of objects under varying illumination conditions
AU2003301143A1 (en) 2002-12-19 2004-07-14 Drexel University Apparatus and method for removing non-discriminatory indices of an indexed dataset
JP4112968B2 (en) * 2002-12-26 2008-07-02 富士通株式会社 Video text processing device
EP1584048A1 (en) * 2003-01-06 2005-10-12 Koninklijke Philips Electronics N.V. Method and apparatus for similar video content hopping
US7194035B2 (en) 2003-01-08 2007-03-20 Apple Computer, Inc. Method and apparatus for improved coding mode selection
US7327885B2 (en) * 2003-06-30 2008-02-05 Mitsubishi Electric Research Laboratories, Inc. Method for detecting short term unusual events in videos
US7403302B2 (en) 2003-08-06 2008-07-22 Hewlett-Packard Development Company, L.P. Method and a system for indexing and tracking digital images
EP1513350A1 (en) 2003-09-03 2005-03-09 Thomson Licensing S.A. Process and arrangement for encoding video pictures
WO2005031654A1 (en) 2003-09-30 2005-04-07 Koninklijke Philips Electronics, N.V. System and method for audio-visual content synthesis
US7313269B2 (en) 2003-12-12 2007-12-25 Mitsubishi Electric Research Laboratories, Inc. Unsupervised learning of video structures in videos using hierarchical statistical models to detect events
US20050154987A1 (en) 2004-01-14 2005-07-14 Isao Otsuka System and method for recording and reproducing multimedia
GB2429597B (en) 2004-02-06 2009-09-23 Agency Science Tech & Res Automatic video event detection and indexing
US7603370B2 (en) 2004-03-22 2009-10-13 Microsoft Corporation Method for duplicate detection and suppression
CN101031882B (en) * 2004-06-08 2010-09-08 达尔特设备互操作有限公司 Architecture, apparatus and method for device team recruitment and content renditioning for universal device interoperability platform
US7590589B2 (en) 2004-09-10 2009-09-15 Hoffberg Steven M Game theoretic prioritization scheme for mobile ad hoc networks permitting hierarchal deference
US7519217B2 (en) * 2004-11-23 2009-04-14 Microsoft Corporation Method and system for generating a classifier using inter-sample relationships
US7308443B1 (en) 2004-12-23 2007-12-11 Ricoh Company, Ltd. Techniques for video retrieval based on HMM similarity
CN101484869B (en) 2005-01-05 2014-11-26 希尔克瑞斯特实验室公司 Scaling and layout methods and systems for handling one-to-many objects
WO2006096612A2 (en) 2005-03-04 2006-09-14 The Trustees Of Columbia University In The City Of New York System and method for motion estimation and mode decision for low-complexity h.264 decoder
US7809192B2 (en) 2005-05-09 2010-10-05 Like.Com System and method for recognizing objects from images and identifying relevancy amongst images and information
JP2008546232A (en) * 2005-05-11 2008-12-18 プラネツトワイド・ゲームズ・インコーポレイテツド Creating publications that use game-based media content
EP1938249A2 (en) 2005-09-02 2008-07-02 Blindsight, Inc. A system and method for detecting text in real-world color images
US7801893B2 (en) * 2005-09-30 2010-09-21 Iac Search & Media, Inc. Similarity detection and clustering of images
US8874477B2 (en) * 2005-10-04 2014-10-28 Steven Mark Hoffberg Multifactorial optimization system and method
US7773813B2 (en) 2005-10-31 2010-08-10 Microsoft Corporation Capture-intention detection for video content analysis
US7644373B2 (en) 2006-01-23 2010-01-05 Microsoft Corporation User interface for viewing clusters of images
US7439989B2 (en) 2006-02-17 2008-10-21 Microsoft Corporation Detecting doctored JPEG images
US7860317B2 (en) 2006-04-04 2010-12-28 Microsoft Corporation Generating search results based on duplicate image detection
US7720851B2 (en) 2006-05-16 2010-05-18 Eastman Kodak Company Active context-based concept fusion
US9049408B2 (en) * 2006-09-01 2015-06-02 Texas Instruments Incorporated Color space appearance model video processor
US7756338B2 (en) 2007-02-14 2010-07-13 Mitsubishi Electric Research Laboratories, Inc. Method for detecting scene boundaries in genre independent videos
US8230457B2 (en) 2007-03-07 2012-07-24 The Nielsen Company (Us), Llc. Method and system for using coherence of biological responses as a measure of performance of a media
WO2008146616A1 (en) 2007-05-25 2008-12-04 Nec Corporation Image-sound section corresponding device, method and program
JP4632096B2 (en) * 2007-06-07 2011-02-16 ソニー株式会社 Navigation device and nearest search method
US7996762B2 (en) 2007-09-21 2011-08-09 Microsoft Corporation Correlative multi-label image annotation
WO2009126785A2 (en) * 2008-04-10 2009-10-15 The Trustees Of Columbia University In The City Of New York Systems and methods for image archaeology
WO2009155281A1 (en) 2008-06-17 2009-12-23 The Trustees Of Columbia University In The City Of New York System and method for dynamically and interactively searching media data
US8671069B2 (en) 2008-12-22 2014-03-11 The Trustees Of Columbia University, In The City Of New York Rapid image annotation via brain state decoding and visual pattern mining
WO2010075408A1 (en) 2008-12-22 2010-07-01 The Trustees Of Columbia University In The City Of New York System and method for annotating and searching media
US8135221B2 (en) * 2009-10-07 2012-03-13 Eastman Kodak Company Video concept classification using audio-visual atoms

Cited By (41)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US11036992B2 (en) 2008-11-17 2021-06-15 Liveclips Llc Method and system for segmenting and transmitting on-demand live-action video in real-time
US11625917B2 (en) 2008-11-17 2023-04-11 Liveclips Llc Method and system for segmenting and transmitting on-demand live-action video in real-time
US9141859B2 (en) 2008-11-17 2015-09-22 Liveclips Llc Method and system for segmenting and transmitting on-demand live-action video in real-time
US9141860B2 (en) 2008-11-17 2015-09-22 Liveclips Llc Method and system for segmenting and transmitting on-demand live-action video in real-time
US10565453B2 (en) 2008-11-17 2020-02-18 Liveclips Llc Method and system for segmenting and transmitting on-demand live-action video in real-time
US10102430B2 (en) 2008-11-17 2018-10-16 Liveclips Llc Method and system for segmenting and transmitting on-demand live-action video in real-time
US9665824B2 (en) 2008-12-22 2017-05-30 The Trustees Of Columbia University In The City Of New York Rapid image annotation via brain state decoding and visual pattern mining
US10491961B2 (en) 2012-04-24 2019-11-26 Liveclips Llc System for annotating media content for automatic content understanding
US10056112B2 (en) 2012-04-24 2018-08-21 Liveclips Llc Annotating media content for automatic content understanding
US9659597B2 (en) 2012-04-24 2017-05-23 Liveclips Llc Annotating media content for automatic content understanding
US9367745B2 (en) 2012-04-24 2016-06-14 Liveclips Llc System for annotating media content for automatic content understanding
US10381045B2 (en) 2012-04-24 2019-08-13 Liveclips Llc Annotating media content for automatic content understanding
US10553252B2 (en) 2012-04-24 2020-02-04 Liveclips Llc Annotating media content for automatic content understanding
US9607053B2 (en) * 2012-08-22 2017-03-28 Expert System France Methods and systems for searching and displaying a plurality of entities within an interactive user interface
US20140059039A1 (en) * 2012-08-22 2014-02-27 Kirk Baker Methods and systems for searching and displaying a plurality of entities within an interactive user interface
WO2015126830A1 (en) * 2014-02-21 2015-08-27 Liveclips Llc System for annotating media content for automatic content understanding
EP3110165A4 (en) * 2014-04-11 2017-08-09 Samsung Electronics Co., Ltd. Broadcast receiving apparatus and method for summarized content service
RU2669574C2 (en) * 2014-04-11 2018-10-12 Самсунг Электроникс Ко., Лтд. Broadcast receiving apparatus and method for summarized content service
US20170286777A1 (en) * 2014-09-04 2017-10-05 Intel Corporation Real time video summarization
US10755105B2 (en) 2014-09-04 2020-08-25 Intel Corporation Real time video summarization
US9639762B2 (en) 2014-09-04 2017-05-02 Intel Corporation Real time video summarization
WO2016036460A1 (en) * 2014-09-04 2016-03-10 Intel Corporation Real time video summarization
US9330084B1 (en) 2014-12-10 2016-05-03 International Business Machines Corporation Automatically generating question-answer pairs during content ingestion by a question answering computing system
CN108140032A (en) * 2015-10-28 2018-06-08 英特尔公司 Automatic video frequency is summarized
US9818032B2 (en) 2015-10-28 2017-11-14 Intel Corporation Automatic video summarization
WO2017074562A1 (en) * 2015-10-28 2017-05-04 Intel Corporation Automatic video summarization
US10157326B2 (en) 2015-10-30 2018-12-18 Xiaomi Inc. Method and device for character area identification
RU2639668C2 (en) * 2015-10-30 2017-12-21 Сяоми Инк. Method and device for region identification
WO2019195835A1 (en) * 2018-04-06 2019-10-10 Deluxe One Llc Comparing frame data to generate a textless version of a multimedia production
US11010627B2 (en) 2019-01-25 2021-05-18 Gracenote, Inc. Methods and systems for scoreboard text region detection
WO2020154555A1 (en) * 2019-01-25 2020-07-30 Gracenote, Inc. Methods and systems for scoreboard text region detection
US11036995B2 (en) 2019-01-25 2021-06-15 Gracenote, Inc. Methods and systems for scoreboard region detection
US10997424B2 (en) 2019-01-25 2021-05-04 Gracenote, Inc. Methods and systems for sport data extraction
US11087161B2 (en) 2019-01-25 2021-08-10 Gracenote, Inc. Methods and systems for determining accuracy of sport-related information extracted from digital video frames
EP3915243A4 (en) * 2019-01-25 2022-10-12 Gracenote Inc. Methods and systems for scoreboard region detection
US11568644B2 (en) 2019-01-25 2023-01-31 Gracenote, Inc. Methods and systems for scoreboard region detection
WO2020154554A1 (en) 2019-01-25 2020-07-30 Gracenote, Inc. Methods and systems for scoreboard region detection
US11792441B2 (en) 2019-01-25 2023-10-17 Gracenote, Inc. Methods and systems for scoreboard text region detection
US11798279B2 (en) 2019-01-25 2023-10-24 Gracenote, Inc. Methods and systems for sport data extraction
US11805283B2 (en) 2019-01-25 2023-10-31 Gracenote, Inc. Methods and systems for extracting sport-related information from digital video frames
US11830261B2 (en) 2019-01-25 2023-11-28 Gracenote, Inc. Methods and systems for determining accuracy of sport-related information extracted from digital video frames

Also Published As

Publication number Publication date
WO2003051031A3 (en) 2004-03-04
US20040255249A1 (en) 2004-12-16
US8488682B2 (en) 2013-07-16
WO2003051031A2 (en) 2003-06-19
AU2002351310A8 (en) 2003-06-23
US20080303942A1 (en) 2008-12-11
US7339992B2 (en) 2008-03-04
AU2002351310A1 (en) 2003-06-23

Similar Documents

Publication Publication Date Title
US8488682B2 (en) System and method for extracting text captions from video and generating video summaries
JP4643829B2 (en) System and method for analyzing video content using detected text in a video frame
Yu et al. Trajectory-based ball detection and tracking with applications to semantic analysis of broadcast soccer video
JP5420199B2 (en) Video analysis device, video analysis method, digest automatic creation system and highlight automatic extraction system
US8971585B2 (en) Image processing apparatus for retrieving object from moving image and method thereof
Zhang et al. General and domain-specific techniques for detecting and recognizing superimposed text in video
JP2002232840A (en) Method for summarizing moving picture
Li et al. Automatic detection and analysis of player action in moving background sports video sequences
Oh et al. Content-based scene change detection and classification technique using background tracking
Sevilmiş et al. Automatic detection of salient objects and spatial relations in videos for a video database system
US20070061727A1 (en) Adaptive key frame extraction from video data
Duan et al. Semantic shot classification in sports video
Mei et al. Structure and event mining in sports video with efficient mosaic
Mei et al. Sports video mining with mosaic
US20070124678A1 (en) Method and apparatus for identifying the high level structure of a program
JP3379453B2 (en) Caption region detection method and device, and moving image search method and device
Jiang et al. Player detection and tracking in broadcast tennis video
Shih et al. Content extraction and interpretation of superimposed captions for broadcasted sports videos
Kim et al. Extracting semantic information from basketball video based on audio-visual features
Lin et al. A feature-based Scheme for detecting and classifying video-shot transitions based on spatio-temporal analysis and fuzzy classification
EP1320992A2 (en) Method for highlighting important information in a video program using visual cues
Han et al. Key Frame Extraction Based on Sports Statistical Characteristics
Wang et al. A motion-insensitive dissolve detection method with SURF
Sang et al. Rolling and non-rolling subtitle detection with temporal and spatial analysis for news video
Rong et al. Gradual transition detection using em curve fitting

Legal Events

Date Code Title Description
STCB Information on status: application discontinuation

Free format text: ABANDONED -- FAILURE TO RESPOND TO AN OFFICE ACTION