CN102103686B - Video identification system using symmetric information of graded image block and method thereof - Google Patents
Video identification system using symmetric information of graded image block and method thereof Download PDFInfo
- Publication number
- CN102103686B CN102103686B CN201010246670.3A CN201010246670A CN102103686B CN 102103686 B CN102103686 B CN 102103686B CN 201010246670 A CN201010246670 A CN 201010246670A CN 102103686 B CN102103686 B CN 102103686B
- Authority
- CN
- China
- Prior art keywords
- video
- frame
- piece
- symmetric information
- feature
- 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.)
- Expired - Fee Related
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/90—Details of database functions independent of the retrieved data types
- G06F16/907—Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/70—Information retrieval; Database structures therefor; File system structures therefor of video data
- G06F16/78—Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually
- G06F16/783—Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually using metadata automatically derived from the content
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T3/00—Geometric image transformation in the plane of the image
- G06T3/40—Scaling the whole image or part thereof
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N1/00—Scanning, transmission or reproduction of documents or the like, e.g. facsimile transmission; Details thereof
- H04N1/40—Picture signal circuits
- H04N1/40012—Conversion of colour to monochrome
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
- H04N19/10—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding
- H04N19/102—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding characterised by the element, parameter or selection affected or controlled by the adaptive coding
- H04N19/119—Adaptive subdivision aspects, e.g. subdivision of a picture into rectangular or non-rectangular coding blocks
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
- H04N19/10—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding
- H04N19/102—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding characterised by the element, parameter or selection affected or controlled by the adaptive coding
- H04N19/12—Selection from among a plurality of transforms or standards, e.g. selection between discrete cosine transform [DCT] and sub-band transform or selection between H.263 and H.264
- H04N19/122—Selection of transform size, e.g. 8x8 or 2x4x8 DCT; Selection of sub-band transforms of varying structure or type
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
- H04N19/10—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding
- H04N19/102—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding characterised by the element, parameter or selection affected or controlled by the adaptive coding
- H04N19/124—Quantisation
- H04N19/126—Details of normalisation or weighting functions, e.g. normalisation matrices or variable uniform quantisers
Landscapes
- Engineering & Computer Science (AREA)
- Multimedia (AREA)
- Theoretical Computer Science (AREA)
- Signal Processing (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Library & Information Science (AREA)
- Databases & Information Systems (AREA)
- Data Mining & Analysis (AREA)
- General Engineering & Computer Science (AREA)
- Discrete Mathematics (AREA)
- Image Analysis (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The present invention discloses a video identification system which uses symmetric information of graded image block and a method thereof. When video clips are input, a frame rate of a video signal is converted to a preset value so the video signal is robust relatively to the conversion in a time axis. Afterwards, a gray grade conversion is performed for only using the brightness information of the video signal. Then, a frame size of the video signal is standardized to a preset size so the video is robust relatively to the size conversion. The frame size is divided to graded blocks by the standardized video, and furthermore symmetric information is extracted from each block thereby generating a specific vector. The graded blocks can be defined along a time axis or along a symmetric structure at a random position in the space for obtaining a graded structure on time or on time-space.
Description
The cross reference of related application
The application advocates to enjoy in the right of priority of the korean patent application that on Dec 21st, 2009 is 10-2009-0127713 to the application number of Department of Intellectual Property of Korea S submission according to 35U.S.C. § 119, whole disclosures of described application case are incorporated herein by reference.
Technical field
Following discloses relate to a kind of video recognition system and method, and are particularly related to a kind of video recognition system and method for the symmetric information that uses classification image block.
Background technology
The variation of terminal, the realization of large-capacity storage media and high-speed communication environment make digitized content be easy to be played, and by fast transport and shared.In addition, due to digitized feature, be easy to transmission and shared the illegal content having with original contents same quality, therefore the infringement of copyright increases.
Therefore, carry out copyright protection to prevent illegal need to the increasing of sharing of large capacity and high-quality video content.For copyright protection, the demand of video monitoring and filtering system is increased.This video monitoring and filtering system are extracted unique video features information (also referred to as " content DNA ") from need the original video of copyright protection; this characteristic information is stored in database (DB); in the time of transmission or shared video content, from video content, extract video features information; extracted information and the information being stored in DB are compared, and result is carried out monitoring and is filtered based on the comparison.
For this video monitoring and filtering system, importantly extract the video features at operation robusts (robust) such as transmission or contingent compression when shared video, size conversion, frame-rate conversion.Especially, recently there are the needs of processing following content, due to the content of supporting as the option of video player to change with 90 ° of rotations or reverse mode.
Under this background, carry out the much research to video identification, and proposed following prior art.
Exercise question is that the Korean Patent No.10-0644016 of " moving image search system and method " (" moving image search system and method ") has proposed a kind of use system that wherein image section of occurrence scene (or camera lens (shot)) change and annotation, color, shape and the texture information of image are searched for video in moving image.But, for by this system applies in filtering, the operation that need to analyze and explain video, and therefore will take a long time to configure DB for large-capacity video.In addition, be difficult to ensure the objectivity of annotation.In addition, because the image of scene change part may be easy to change due to factors such as frame-rate conversion, so the reliability of search may worsen.
Exercise question has proposed a kind of scene change for detection of vision signal and has utilized length between scene change to identify the scheme of video for the Korean Patent No.10-0729660 of digital video recognition system and the method for length " use scenes change " (" digital video identification system and method using scene change length ").In the time identifying video based on scene change, scene change quantity, and thereby may go wrong when DB when configuration or search for very large or very little according to video to be searched.
The exercise question of Job Oosteven, Ton Kalker and Japp Haitsma is the video frequency identifying method that the paper of " for feature extraction and the database policies (Feature Extraction and a Database Strategy for Video Fingerprinting) (Proceeding of International Conference on Recent Advances in Visual Information Systems, 2002) of video fingerprint recognition " has proposed a kind of brightness value based on image block.In this section of paper, obtain the average brightness value of image block, extract feature with the time between brightness value and spatial diversity.In this case, because feature is by binarization, so can improve search efficiency.But, because used the difference of the continuous blocks with unified size, so can not identify rotation, image reversion and distortion, in the efficiency that relates to large capacity DB application, search time etc., go wrong.
In the low computational load of needs, must be robust for monitoring with the video identification technology of filtering system for operations such as the size conversion occurring when transmission or the shared video, compression, frame-rate conversion, rotation, reversions.In addition, feature can not depend on the feature based on style (genre-based characteristics) of video.For example, between action movie and feature film, can not there is the difference on discrimination, wherein in action movie, have a large amount of motions and scene change to occur, and in feature film, only have the scene change of relatively small amount or motion to occur.
Summary of the invention
Correspondingly, consider the problems referred to above that occur in prior art and make the present invention, and the object of the invention is to provide a kind of video recognition system and method, this system and method is in extracting video features with low computational load, for being robust in transmission or the various distortions that may occur when shared video, for being robust owing to the distortion of size conversion, compression, frame-rate conversion, rotation and reversion.
To achieve these goals, the invention provides a kind of video recognition system, comprise: feature and metamessage database (DB) unit, for store extract from multiple video clippings and many video informations and be the needed feature of video identification and video element information; Feature extraction unit, for extracting feature from input video montage; Database search unit, for coming search characteristics and metamessage database by extracted feature; And characteristic matching unit, for the Search Results of extracted feature and described feature and metamessage database is matched; Wherein feature extraction unit comprises: frame-rate conversion unit, for the frame-rate conversion of the vision signal of input video montage is become to a preset value; Grey level transition unit, carries out grey level transition for the vision signal that frame rate has been converted; Frame sign standardized unit, for being normalized to the frame sign of the vision signal of executed grey level transition default size; And blocking characteristic extraction unit, for being classification piece by Video segmentation, extract symmetric information from each piece, and generating feature vector then.
Preferably, blocking characteristic extraction unit can become classification piece by Video segmentation by the piece pattern based on having time or space symmetr structure.More specifically, blocking characteristic extraction unit based on being defined as space symmetr structure, simultaneously in successive frame, there is time hierarchy, there is in time the time-space classification piece pattern of time hierarchy simultaneously in different frame, be classification piece by Video segmentation.Blocking characteristic extraction unit can generate the eigenvector that comprises N dimension symmetric information eigenwert based on piece pattern.
Preferably, database search unit can be by changing the positional value of N dimension symmetric information eigenwert or coming search characteristics and metamessage database by inverse characteristic value, to determine because of orthogonal rotation or the horizontal/vertical distortion video causing that reverses, and can be only with N dimension symmetric information eigenwert come tentatively search characteristics and metamessage database compared with upper strata bit, and secondly only search for preliminary Search Results by the remaining bits of N dimension symmetric information eigenwert.
To achieve these goals, the invention provides a kind of method of extracting feature from video, comprising: by being that preset value is carried out frame-rate conversion by the frame-rate conversion of incoming video signal; The vision signal that frame rate has been converted is carried out grey level transition; By being normalized to, the frame sign of the vision signal of executed grey level transition specify size to carry out frame sign normalization; And by by Video segmentation be classification piece, by from each piece, extract symmetric information and by generating feature vector carry out blocking characteristic extract.
In addition, the invention provides a kind of video frequency identifying method, comprising: by being that preset value is carried out frame-rate conversion by the frame-rate conversion of incoming video signal; The vision signal that frame rate has been converted is carried out grey level transition; By being normalized to, the frame sign of the vision signal of executed grey level transition specify size to carry out frame sign normalization; By being classification piece by Video segmentation, extracting symmetric information and generating feature vector carry out blocking characteristic and extract from each piece; And by searching for the property data base (DB) of use characteristic vector configuration in advance and the Search Results of eigenvector and property data base being mated, carry out characteristic matching.
According to detail specifications, accompanying drawing, claims below, other features and aspect are by apparition.
Brief description of the drawings
Fig. 1 shows according to the block diagram of the structure of the video recognition system of the symmetric information of the use classification image block of the embodiment of the present invention.
Fig. 2 shows according to the block diagram of the structure of the feature extraction unit of the video recognition system of the embodiment of the present invention.
Fig. 3 shows the figure of the example of the characteristic extraction procedure of being carried out by the feature extraction unit of Fig. 2.
Fig. 4 a and Fig. 4 b are the figure that shows respectively the example of the image block with time-space hierarchy.
Fig. 5 shows the piece pattern based in Fig. 4 a and Fig. 4 b and the figure of the example of the N dimensional feature value extracted.
Fig. 6 shows the figure of the example of the bit operating for calculating similarity.
Embodiment
Hereinafter, detailed description exemplary embodiment with reference to the accompanying drawings.Throughout the drawings and detailed description, unless otherwise described, identical Reference numeral will be understood to refer to identical element, feature and structure.For clear, explanation and convenient for the purpose of, the relative size of these elements and narration may be exaggerated.Following detailed description obtains the complete understanding to method described herein, equipment and/or system with helping reader.Thereby various variations, amendment and the equivalent of method described herein, equipment and/or system will be that those of ordinary skill in the art institute is thinkable.And for further clear and concise and to the point, the description of known function and structure may be omitted." comprise (include) ", implication specified properties, region, fixed number, step, technique, element and/or the composition of " comprising (comprise) ", " comprising (including) " or " comprising (comprising) ", but do not get rid of other character, region, fixed number, step, technique, element and/or composition.
Hereinafter, detailed description exemplary embodiment with reference to the accompanying drawings.
According to the present invention, by using the symmetric information of classification image block to extract the identifying information (, the feature of content) of content from vision signal, identify content.For this operation, from any classification piece of vision signal, obtain symmetric information, arbitrarily classification piece value of being subdivided into, these values are configured to the form of matrix, and entry of a matrix element is used as eigenwert, thus video is identified.
Fig. 1 shows according to the block diagram of the structure of the video recognition system of the symmetric information of the use classification image block of the embodiment of the present invention.
As shown in Figure 1, comprise feature and metamessage database (DB) 110, feature extraction unit 120, DB search unit 130 and characteristic matching unit 140 according to the video recognition system 100 of the symmetric information of the use classification image block of the embodiment of the present invention.
By extract feature (content DNA) and the video element information for video identification by multiple video clippings and multi-disc video information, come pre-configured feature and metamessage DB 110.
Feature extraction unit 120 is the assemblies for extracting the feature of expecting the video clipping being identified, and will describe afterwards its structure and function in detail.DB search unit 130 features that use is extracted are come search characteristics and metamessage DB 110, and characteristic matching unit 140 matches the Search Results of extracted feature and DB.By this mode, can obtain the information about input video montage.
According in the video recognition system of the symmetric information of the use classification image block of the embodiment of the present invention, use the feature of the symmetric information based on classification image block as the eigenwert of video.The structure that is used for the feature extraction unit of extracting these features is illustrated in Fig. 2.Fig. 3 illustrates the example of the characteristic extraction procedure of being carried out by feature extraction unit.
As shown in Figure 2, comprise frame-rate conversion unit 121, grey level transition unit 123, frame sign standardized unit 125 and piecemeal (block-wise) feature extraction unit 127 according to the feature extraction unit 120 in video recognition system 100 of the present invention.
As shown in Figure 3, in the time of video clipping 320 that input comprises multiframe, the vision signal frame-rate conversion of input video montage is become a preset value by frame-rate conversion unit 121, thereby and to make vision signal be robust for the conversion that may occur on time shaft.For example, no matter the frame rate of incoming video signal how, it is the conversion operations of the same frame rate preset that frame-rate conversion unit 121 is all carried out the frame-rate conversion of incoming video signal.As the result of conversion, generate the vision signal 321 with default same frame rate.
By grey level transition unit 123, vision signal is carried out to grey level transition.Grey level transition unit 123 is carried out for vision signal is converted to the process of grayscale image, thereby only has the monochrome information of vision signal to be used, and the colouring information of vision signal can be left in the basket.As the result of conversion, generate gray scale video signal 323.
Next, the frame sign of vision signal is normalized to default size by frame sign standardized unit 125, is robust to make this video for size conversion.Therefore, having generated its frame sign has been normalized as default big or small vision signal 324.
Finally, blocking characteristic extraction unit 127 is cut apart frame to make video be divided into classification piece 325, from each piece, extracts symmetric information, and generating feature vector afterwards.
The image block of each video can be by being defined along time shaft or along the symmetrical structure of optional position, space.The example with the image block of this symmetrical structure is illustrated in Fig. 4 a and Fig. 4 b.Fig. 4 a illustrates time hierarchy, and Fig. 4 b illustrates spatial scalability structure.To the structure of these pieces be described below.
From four pieces of constructing as shown in the figure, obtain symmetric information.In this situation, as shown in Figure 4 b, can select four pieces with symmetrical structure to obtain spatial scalability structure, or in spatial scalability structure, obtain time-space (temporal-spatial) hierarchy by the time hierarchy in Fig. 4 a is applied to.When having the piece of time-space hierarchy when selected, the one or more time hierarchies in Fig. 4 a can be applied in spatial scalability structure.
The process of the symmetric information for extracting four pieces is described below.First,, in the time there is 2 × 2 matrix A as shown in equation (1), can obtain by following equation (2) symmetric information of matrix A.
Even if video is rotated or is inverted with 90 ° of angles, the value of this symmetric information is still kept.For example, in the time that video is rotated 90 ° of angles, provide matrix A ' in the situation that by following equation (3), matrix A ' symmetric information can obtain by following equation (4).
By with the method that method is identical above, even for the matrix A along Z-axis reversion as shown in equation (5) ", can obtain the result of equation (6).
S
1(A″)＝S
1(A)，S
2(A″)＝～S
2(A)，S
3(A″)＝～S
3(A) (6)
That is to say, even if video is rotated 90 ° of angles, also only changed the position of the feature of each piece, (for example, the S and eigenwert remains unchanged
1(A ')=～S
2(A)).In addition, though by 180 ° or 270 ° of rotations together with 90 ° of angles rotations and level or vertically reverse and cause distortion in the situation that, eigenwert still remains unchanged.
Because these features, according to the symmetric information eigenwert of the embodiment of the present invention for the horizontal or vertical reversion by video or be robust with the distortion that 90 °, 180 ° or 270 ° of angle rotating videos cause.Particularly, in the situation of 90 ° and 270 ° rotations, the position of eigenwert changes, but video is to be that therefore position can be accurate on the position having changed in N dimensional feature position based on recently the determining in length and breadth of video with 90 ° or 270 ° of angle rotations.
According to the embodiment of the present invention, by repeat to have time-space classification feature subordinate's piece but not arbitrarily piece build four the required pieces of eigenvector that formed by blocking characteristic for generating.That is to say, as shown in Fig. 4 a and Fig. 4 b, definition time-spatial scalability piece pattern (Block Patterns, BP), this piece pattern has spatial scalability structure and in successive frame, also has time hierarchy simultaneously, and thereby definable N dimension symmetric information eigenwert.Here, time hierarchy is also not necessarily defined in successive frame, and four pieces in upper different frame of the time that also can be defined as have symmetrical structure.
Equation (1) can be applied to the piece pattern defining in Fig. 4 a and Fig. 4 b and extract N dimensional feature value, its example as shown in Figure 5.
In the example depicted in fig. 5, BP1 has three-dimensional feature, and S1 (BP1), S2 (BP1) and S3 (BP1) extract from same frame, or in different frame, extract from the time, to obtain the time hierarchy as shown in the example of Fig. 4 a.That is to say, when extract piece pattern (BP) from same number of frames time, only usage space hierarchy.In the time extracting piece pattern in frames different from the time, service time-spatial scalability structure.
Conventionally from being robust as the symmetric information eigenwert of extracting compared with wide pattern on upper strata for video distortion, but in identification, be fragile (that is to say, have the image much with same characteristic features value).In addition, there is high sense (that is to say, have the image seldom with same characteristic features value) from the symmetric information eigenwert of extracting as narrow pattern of lower level, but be fragile for video distortion.Consistent with the feature of these eigenwerts, obtain and use compared with characteristic classification candidate's video group on upper strata and use the residue character of lower level to identify the advantage of video.
That is to say, in with reference to the described characteristic matching process of figure 1, the feature of not mutual more all dimensions, but in the feature group of classifying compared with the feature on upper strata (or attribute) in use, the feature that mutually compares lower level, has therefore improved search speed.
In embodiments of the present invention, use the characteristic information of classification symmetric information value as video, and extract symmetric information value from the prescribed fractionated piece of normalization frame.For example, when hypothesis frame rate is 10fps, the size of normalized images is 8*8, and the quantity of spatial scalability piece pattern is 18 o'clock, and the video clipping of about 10 seconds has for the eigenwert of 5400 bits of 10 seconds * 10fps=100 frames (18 * 3 every frame=54 of symmetry dimension/frames).In addition, in the time that the quantity of spatial scalability piece pattern is confirmed as 18, and temporal scalability piece patterning is only applied to present frame and proper that frame before present frame, can be further extract 18 patterns with the quantity of temporal scalability piece pattern each frame pro rata, from 99 frames of 100 frames altogether.
Configure by above-mentioned eigenwert the feature DB that search system will be used in advance, and this feature DB is used from search with input metamessage DB mono-.In search system, use following equation (7) to carry out the distance between comparative feature.For example, when the eigenwert of the i frame in the N dimensional feature that hypothesis receives as search input (query excerpt) is Q (i), and the eigenwert of the i frame of k video in DB is DB (k, i), time, provide as follows the algorithm for query excerpt and DB are compared.
Fig. 6 shows the figure of the example of the bit operating for calculating similarity.As shown in Figure 6, when carrying out after the computing of XNOR bit to N dimensional feature value, when counting 1, the similarity value (namely, similarity distance D) of measurement can be calculated by following equation (7).
The definite of similarity (S) with the query excerpt of frame length m is to adopt in such a way to carry out, in the time that the similarity value of the successive frame of measuring is greater than predetermined threshold, determine that query excerpt is the video being stored in DB, as shown in equation (8).
In addition,, in order to distinguish the distortion video causing because of orthogonal rotation (90 °, 180 ° and 270 °), horizontal/vertical reversion etc., only need position or inverse characteristic value by changing N dimensional feature to carry out aforesaid operations.
In addition, only with compared with upper strata bit, DB preliminary classification being become to similar group in N dimensional feature, and only in sorted group, carry out search, thereby improved search performance.
In search system, as recognition result, the position of the i frame of output similarity k video that be maximized, in DB.
Therebetween, in an embodiment of the present invention, describe difference between the classification symmetrical structure that adopts video and extracted feature use characteristic and identify the method and system of video, but it is apparent that, also can be applicable to the field except video identification according to the Feature Extraction System of the embodiment of the present invention and method.
Can be embodied as the computer-readable code on computer readable recording medium storing program for performing according to the video frequency identifying method of the symmetric information of the use classification image block of the embodiment of the present invention.Computer-readable recording medium is that on it, stored can be by the pen recorder of the data of computer system reads, and can be for example ROM (read-only memory) (ROM), random access memory (RAM), cache memory, hard disk, flexible plastic disc, flash memory or optical data storage device.In addition, described medium can provide with carrier format, and for example can be included in situation about providing on the Internet.In addition, computer-readable recording medium can be distributed in the computer system connected to one another on network, and can store and carry out using distributed way as computer readable code.
As mentioned above, advantage of the present invention is: simplify the recognition property (or feature) of digital video with the time-space hierarchy of digital video, thereby improved search performance.In addition, based on according to video and contingent various block size local wrong and hierarchical nature inversely proportional the fact that causes, only have the attribute on upper strata can be used to index and classification.In addition, by only changing the position of the attribute dimension of the attribute that depends on video recognition system, just can search for discernible distortion environment (rotation, reversion etc.).In addition can control by the amplitude that changes attribute dimension recognition speed and search time.
Although for illustration purpose discloses according to the video recognition system of symmetric information of use classification image block of the present invention and the preferred embodiment of method, but those skilled in the art can understand, in the situation that not deviating from scope and spirit of the present invention disclosed in the accompanying claims, may carry out various amendments, increase and replacement.
Claims (20)
1. a video recognition system, comprising:
Feature and metamessage database;
Be used for by becoming a preset value to carry out the device of frame-rate conversion the frame-rate conversion of incoming video signal;
Carry out the device of grey level transition for the vision signal that frame rate has been converted;
For specifying size by the frame sign of the vision signal of executed grey level transition is normalized to, carrying out the normalized device of frame sign;
For by being to have along time shaft or along the classification piece of the symmetrical structure of optional position, space, extract symmetric information and generating feature vector is carried out the device that blocking characteristic extracts from each piece Video segmentation; And
For the feature by search use characteristic vector configuration in advance and metamessage database and the Search Results of eigenvector and feature and metamessage database is mated to carry out the device of characteristic matching.
2. video recognition system according to claim 1, the piece pattern of the device that wherein said execution blocking characteristic extracts based on having space symmetr structure becomes to have along the classification piece of the symmetrical structure of locus by Video segmentation.
3. video recognition system according to claim 2, the device that wherein said execution blocking characteristic extracts generates the eigenvector that comprises N dimension symmetric information eigenwert based on piece pattern.
4. video recognition system according to claim 3, the device of wherein said execution characteristic matching is by changing the positional value of N dimension symmetric information eigenwert or coming search characteristics and metamessage database by inverse characteristic value, to determine because of orthogonal rotation or the horizontal/vertical distortion video causing that reverses.
5. video recognition system according to claim 3, the device of wherein said execution characteristic matching only comes tentatively search characteristics and metamessage database with the high order bit of N dimension symmetric information eigenwert, and secondly only searches for preliminary Search Results by the remaining bits of N dimension symmetric information eigenwert.
6. video recognition system according to claim 1, the piece pattern of the device that wherein said execution blocking characteristic extracts based on have space symmetr structure and have in different frame in time time hierarchy simultaneously, is classification piece by Video segmentation.
7. video recognition system according to claim 6, the device that wherein said execution blocking characteristic extracts generates the eigenvector that comprises N dimension symmetric information eigenwert based on piece pattern.
8. video recognition system according to claim 7, the device of wherein said execution characteristic matching is by changing the positional value of N dimension symmetric information eigenwert or coming search characteristics and metamessage database by inverse characteristic value, in order to determine because of orthogonal rotation or the horizontal/vertical distortion video causing that reverses.
9. video recognition system according to claim 7, the device of wherein said execution characteristic matching only comes tentatively search characteristics and metamessage database with the high order bit of N dimension symmetric information eigenwert, and secondly uses N to tie up the remaining bits of symmetric information eigenwert and only search for preliminary Search Results.
10. a method of extracting feature from video, comprising:
By being that preset value is carried out frame-rate conversion by the frame-rate conversion of incoming video signal;
The vision signal that frame rate has been converted is carried out grey level transition;
By being normalized to, the frame sign of the vision signal of executed grey level transition specify size to carry out frame sign normalization; And
By being to have along time shaft or along the classification piece of the symmetrical structure of optional position, space, by extracting symmetric information and carry out blocking characteristic by generating feature vector and extract from each piece Video segmentation.
11. methods according to claim 10, the step that wherein said execution blocking characteristic extracts is configured to based on the piece pattern with space symmetr structure, Video segmentation be become to have along the classification piece of the symmetrical structure of locus.
12. methods according to claim 11, the step that wherein said execution blocking characteristic extracts is configured to generate based on described pattern the eigenvector that comprises N dimension symmetric information eigenwert.
13. methods according to claim 10, the step that wherein said execution blocking characteristic extracts is configured to, based on the piece pattern that has space symmetr structure and have in different frame in time time hierarchy simultaneously, Video segmentation is become to classification piece.
14. methods according to claim 13, the step that wherein said execution blocking characteristic extracts is configured to generate based on described pattern the eigenvector that comprises N dimension symmetric information eigenwert.
15. 1 kinds of video frequency identifying methods, comprising:
By being that preset value is carried out frame-rate conversion by the frame-rate conversion of incoming video signal;
The vision signal that frame rate has been converted is carried out grey level transition;
By being normalized to, the frame sign of the vision signal of executed grey level transition specify size to carry out frame sign normalization;
By being to have along time shaft or along the classification piece of the symmetrical structure of optional position, space, extract symmetric information and generating feature vector carry out blocking characteristic and extract from each piece Video segmentation; And
Feature by search use characteristic vector configuration in advance and metamessage database and eigenvector is mated with the Search Results of described feature and metamessage database, carry out characteristic matching.
16. video frequency identifying methods according to claim 15, it is the classification piece having along the symmetrical structure of locus by Video segmentation that the step that wherein said execution blocking characteristic extracts is configured to based on the piece pattern with space symmetr structure.
17. video frequency identifying methods according to claim 15, the step that wherein said execution blocking characteristic extracts is configured to, based on the piece pattern that has space symmetr structure and have in different frame in time time hierarchy simultaneously, Video segmentation is become to classification piece.
18. video frequency identifying methods according to claim 17, the step that wherein said execution blocking characteristic extracts is configured to generate based on described pattern the eigenvector that comprises N dimension symmetric information eigenwert.
19. video frequency identifying methods according to claim 18, the step of wherein said execution characteristic matching is configured to: by changing the positional value of N dimension symmetric information eigenwert or searching for described feature and metamessage database by inverse characteristic value, to determine by orthogonal rotation or the horizontal/vertical distortion video causing that reverses.
20. video frequency identifying methods according to claim 18, the step of wherein said execution characteristic matching is configured to: only use the high order bit of N dimension symmetric information eigenwert tentatively to search for described feature and metamessage database, and secondly use N to tie up the remaining bits of symmetric information eigenwert and only search for preliminary Search Results.
Applications Claiming Priority (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| KR1020090127713A KR101367821B1 (en) | 2009-12-21 | 2009-12-21 | video identification method and apparatus using symmetric information of hierachical image blocks |
| KR127713/09 | 2009-12-21 |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN102103686A CN102103686A (en) | 2011-06-22 |
| CN102103686B true CN102103686B (en) | 2014-10-01 |
Family
ID=44156442
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201010246670.3A Expired - Fee Related CN102103686B (en) | 2009-12-21 | 2010-06-30 | Video identification system using symmetric information of graded image block and method thereof |
Country Status (2)
| Country | Link |
|---|---|
| KR (1) | KR101367821B1 (en) |
| CN (1) | CN102103686B (en) |
Families Citing this family (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| KR101537032B1 (en) * | 2014-05-15 | 2015-07-15 | 심해청 | Method for generating feature information for image filtering and recording-medium recorded program thereof |
| CN104219423B (en) * | 2014-09-25 | 2017-09-29 | 联想(北京)有限公司 | A kind of information processing method and device |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1893676A (en) * | 2005-06-29 | 2007-01-10 | 株式会社Ntt都科摩 | Video evaluation apparatus, spatio-temporal frequency analyzing apparatus, video evaluation method, spatio-temporal frequency analyzing method, video evaluation program, and spatio-temporal frequency |
| CN101170672A (en) * | 2006-10-24 | 2008-04-30 | 索尼株式会社 | Contents character information acquisition method, contents character information acquisition program, contents character information acquisition apparatus, and video content recording apparatus |
| CN101369281A (en) * | 2008-10-09 | 2009-02-18 | 湖北科创高新网络视频股份有限公司 | Retrieval method based on video abstract metadata |
Family Cites Families (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1238889A (en) | 1996-11-27 | 1999-12-15 | 普林斯顿视频图像公司 | Motion tracking using image-texture templates |
| US6462754B1 (en) | 1999-02-22 | 2002-10-08 | Siemens Corporate Research, Inc. | Method and apparatus for authoring and linking video documents |
| US6879731B2 (en) | 2003-04-29 | 2005-04-12 | Microsoft Corporation | System and process for generating high dynamic range video |
| US8009861B2 (en) * | 2006-04-28 | 2011-08-30 | Vobile, Inc. | Method and system for fingerprinting digital video object based on multiresolution, multirate spatial and temporal signatures |
| US8204314B2 (en) * | 2007-12-03 | 2012-06-19 | Vobile, Inc. | Method and system for fingerprinting digital video object based on multiresolution, multirate spatial and temporal signatures |
| KR100944903B1 (en) * | 2008-03-18 | 2010-03-03 | 한국전자통신연구원 | Feature extraction apparatus of video signal and its extraction method, video recognition system and its identification method |
-
2009
- 2009-12-21 KR KR1020090127713A patent/KR101367821B1/en active IP Right Grant
-
2010
- 2010-06-30 CN CN201010246670.3A patent/CN102103686B/en not_active Expired - Fee Related
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1893676A (en) * | 2005-06-29 | 2007-01-10 | 株式会社Ntt都科摩 | Video evaluation apparatus, spatio-temporal frequency analyzing apparatus, video evaluation method, spatio-temporal frequency analyzing method, video evaluation program, and spatio-temporal frequency |
| CN101170672A (en) * | 2006-10-24 | 2008-04-30 | 索尼株式会社 | Contents character information acquisition method, contents character information acquisition program, contents character information acquisition apparatus, and video content recording apparatus |
| CN101369281A (en) * | 2008-10-09 | 2009-02-18 | 湖北科创高新网络视频股份有限公司 | Retrieval method based on video abstract metadata |
Also Published As
| Publication number | Publication date |
|---|---|
| CN102103686A (en) | 2011-06-22 |
| KR101367821B1 (en) | 2014-02-26 |
| KR20110071209A (en) | 2011-06-29 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US6470094B1 (en) | Generalized text localization in images | |
| US6473522B1 (en) | Estimating text color and segmentation of images | |
| Tokuda et al. | Computer generated images vs. digital photographs: A synergetic feature and classifier combination approach | |
| Sowmya et al. | A survey on video forgery detection | |
| US20150154456A1 (en) | Method and an apparatus for the extraction of descriptors from video content, preferably for search and retrieval purpose | |
| CN103281473B (en) | General video steganalysis method based on video pixel space-time relevance | |
| Doulamis et al. | Non-sequential video content representation using temporal variation of feature vectors | |
| Mizher et al. | A review of video falsifying techniques and video forgery detection techniques | |
| Wary et al. | A review on robust video copy detection | |
| Heng et al. | An object-based shot boundary detection using edge tracing and tracking | |
| Alipour et al. | Semantic segmentation of JPEG blocks using a deep CNN for non-aligned JPEG forgery detection and localization | |
| Gupta et al. | Video authentication in digital forensic | |
| Barbu | Novel automatic video cut detection technique using Gabor filtering | |
| CN102103686B (en) | Video identification system using symmetric information of graded image block and method thereof | |
| Ouyang et al. | The comparison and analysis of extracting video key frame | |
| Agarwal et al. | The advent of deep learning-based image forgery detection techniques | |
| Arai et al. | Text extraction from TV commercial using blob extraction method | |
| Ho et al. | Effective images splicing detection based on decision fusion | |
| Lian et al. | Multimedia forensics for detecting forgeries | |
| Nie et al. | Video fingerprinting based on graph model | |
| Mercier et al. | Detecting manipulations in video | |
| Kumari et al. | Video Object Forgery Detection using Image Processing Techniques | |
| Gan et al. | Video Surveillance Object Forgery Detection using PDCL Network with Residual-based Steganalysis Feature | |
| Jegaveerapandian et al. | A survey on passive digital video forgery detection techniques. | |
| Chauhan et al. | Smart surveillance based on video summarization: a comprehensive review, issues, and challenges |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C14 | Grant of patent or utility model | ||
| GR01 | Patent grant | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20141001 Termination date: 20210630 |
|
| CF01 | Termination of patent right due to non-payment of annual fee |