CN104869403B - A kind of shot segmentation method based on X264 compression videos - Google Patents
A kind of shot segmentation method based on X264 compression videos Download PDFInfo
- Publication number
- CN104869403B CN104869403B CN201510253354.1A CN201510253354A CN104869403B CN 104869403 B CN104869403 B CN 104869403B CN 201510253354 A CN201510253354 A CN 201510253354A CN 104869403 B CN104869403 B CN 104869403B
- Authority
- CN
- China
- Prior art keywords
- frames
- gop
- frame
- shot
- shot segmentation
- 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
Landscapes
- Compression Or Coding Systems Of Tv Signals (AREA)
Abstract
The present invention relates to a kind of shot segmentation methods based on X264 compression videos.The method includes：Frame number, frame type and intra prediction macro block ratio are extracted in decoding frame by frame as unit of the image group GOP of X264 compression videos；Shot Detection is carried out to the I frames in GOP to be measured, judges whether current I frames occur lens mutation according to the length of GOP to be measured；Using the intra prediction macro block ratio of all P frames in the GOP to be measured as sample data, judge whether to need shot segmentation according to the degree of fluctuation of sample data；Shot segmentation is carried out based on predicted macroblock ratio.The present invention can carry out shot segmentation to GOP length and the unfixed compression video of structure.When shot segmentation is happened at I frames, the prior art is generally detected using the variation of predicted macroblock ratio in I frame consecutive frames, it is desirable that video has fixed GOP length and structure.And the shot transition detection to I frames only can be realized in the present invention using the I frames parameters of itself.
Description
Technical field
The invention belongs to video lens to divide field, and in particular to a kind of shot segmentation side based on X264 compression videos
Method.
Background technology
With the development of multimedia, Internet technology, on the one hand, people, which stay indoors, can easily enjoy video point
It broadcasts, various services, the requirement to Video service such as viewing are higher and higher online.On the other hand, media industry is brought for a long time
People's information and amusement, the storage of magnanimity is but also the importance of media asset management becomes increasingly conspicuous.Shot segmentation technique for
More than the solutions of two aspect problems all have great importance.First, shot segmentation technique is based on Video content retrieval
One step is also one of most important step, the effect that accuracy will directly influence subsequent processing.Secondly, it is compiled in media article
In mesh, shot segmentation technique finds out Shot change point automatically according to content deltas, is selected for Category-makers ', substantially increases cataloguing
The degree of automation of system.
Camera lens is the basic unit for forming video, while is also based on the basic unit of the video processing of content.Video counts
It is formed according to by a series of images continuous in time.In the inside of a camera lens, the correlation between image in terms of content
It is stronger, and the correlation between the image in different camera lenses in terms of content is weaker.Shot segmentation is then to the mirror in video
Boundary is detected in front, finds out Shot change point, by Video segmentation into the part more easily handled.Under normal conditions, according to
The speed of camera lens conversion can be classified as mutation conversion and gradual change conversion.
Shot segmentation method can be divided into two major class：Method based on uncompressed domain and the method based on compression domain.It is based on
The shot segmentation method of uncompressed domain needs first to decompress compression video, is then handled again, this is greatly reduced
The efficiency of shot segmentation.Compared with the method based on uncompressed domain, the method based on compression domain has processing without decompressing completely
The advantages that data volume is few, speed is fast, real-time is high, feature extraction facilitates.
Video is compressed mainly for two classes based on the shot segmentation method of compression domain：Mpeg compressed video and H.264/AVC
Compress video.Shot segmentation algorithm based on mpeg compressed video mainly includes the method based on DCT coefficient, the side based on DC figures
Method and method based on motion vector etc..Compared with MPEG compression and coding standards, some differences are H.264/AVC produced,
For example it supports the macroblock partition of different sizes and shapes, support multiple reference frame prediction estimation etc..These new features bring it is excellent
While different compression performance and network applicability, but also these above-mentioned shot segmentation algorithms based on mpeg compressed video are not
It is suitable for H.264/AVC compressing video again.Shot segmentation algorithm based on H.264/AVC compression video is mainly included based on macro block
The method of absolute quantity, the method based on macro block ratio and method based on the distribution of macro block bit etc..Based on macro block absolute quantity
Method the predictive coding pattern of each macro block in frame image is directly extracted on compressed bit stream.Different types of frame is distinguished
Corresponding intraframe predictive coding macroblock number, forward predictive coded macroblock number and back forecast coded macroblocks number are counted, passes through phase
The absolute quantity of different predictive coding mode macros judges shot segmentation with the relationship of preset threshold value between adjacent two frames
The presence of point.Method based on macro block absolute quantity there are one it is apparent the drawbacks of, for the video of different resolution, macro block
Absolute quantity can great changes will take place.Method based on macro block ratio is a kind of improvement based on macro block absolute quantity method,
It is by the way that the percentage using all kinds of predictive coding mode macros counted is compared, and then sentence with preset threshold value
Make the presence of shot segmentation point.Since Shot change has occurred, intra prediction macro block and inter-prediction macro block are in code stream
Bit distribution condition is different.Method based on the distribution of macro block bit is will be used in intra prediction macroblock coding in a frame image
Bit number and its inter-prediction macroblock coding used in bit number ask than being worth to bit distribution condition, then with it is preset
Threshold value is compared, and determines the presence of shot segmentation point.
For H.264/AVC compressing the shot segmentation of video, use the method based on macro block ratio more.Fig. 1 gives two
The schematic diagram of kind shot segmentation situation.
(1) shot segmentation be happened at I frames or P frames (as in Fig. 1 c points, shown in d points)
Since camera lens is converted, there will be larger differences for the I frames or the picture material of P frames and front B frames.At this point, in B frames
Macro block by majority use forward prediction mode.If Shot change is happened at P frames, then the macro block in the P frames adopts majority
Use intra prediction mode.
(2) shot segmentation is happened at B frames
If camera lens conversion is happened at first B frame (as shown in a points in Fig. 1), since camera lens is converted, the B frames and front
I frames or the picture materials of P frames there will be larger differences.Similarly, a frame is similarly B frames behind the B frames, the two B frame categories
In same camera lens.So majority is used back forecast pattern by the macro block in the two B frames.
If camera lens conversion is happened at second B frame (as shown in the b points in Fig. 1), then a frame of front is equally B frames.
Since camera lens is converted, there will be larger differences for the picture material of the two B frames.Therefore the macro block in first B frame adopts majority
With forward prediction mode, and majority is used back forecast pattern by the macro block in second B frame.
The structure of known video GOP (Group of Pictures, image group) before this method requirement shot segmentation.If depending on
The structure of GOP changes when frequency encodes, then this method will be no longer applicable in.In addition, if gradual shot occurs, above-mentioned rule changes
Also unobvious.
Due to being influenced by high complexity, H.264/AVC it is difficult to apply in real time sometimes.And X264 is as H.264/
AVC it is a kind of it is simplified realize, given up some new features H.264/AVC to reduce computation complexity, do not significantly reduced pressure
It is optimized under the premise of contracting performance, to improve efficiency of algorithm.X264 is widely used in practice.
X264 divides GOP according to video content, as shown in Figure 2.Wherein keyint represents the maximum length of GOP, min_
Keyint represents the minimum length of GOP, and visual_difference expressions consider scene difference and computation complexity etc.
Coding cost obtained from aspect, the threshold value of scenecut presentation code costs.The first frame of video is every IDR frame, you can even if
Coded frame, (direction incremental along the display order time) mobile min_keyint distances, visual_ is calculated to present frame backward
difference.If visual_difference is more than scenecut, judges that video scene is changed, that is, be inserted into one
A every IDR frame.If visual_difference is less than scenecut, the current displacement distance of comparison and the relationship of keyint are needed.
If displacement distance reaches keyint, it is inserted into a common I frame.If displacement distance is less than keyint, a frame is moved backward,
Continue the magnitude relationship for comparing visual_difference and scenecut.Similarly, whenever being inserted into an every IDR frame or I frames, then
It needs to move backward min_keyint distances.
The GOP of X264 compression videos does not have fixed length and structure, and conventional compression domain shot segmentation method does not apply to simultaneously.
In addition, X264 provides three kinds of I frames, P frames and B frames frame types.It is found by many experiments, B frames act as carrying to the maximum extent
High compression performance, lens mutation do not occur at B frames.Thus, X264 compression videos need a kind of shot segmentation method of system.
Invention content
It is an object of the present invention to according to the coding principle of X264, the frame type of video, macro block are compressed by extracting X264
The code stream informations such as coding mode provide a kind of shot segmentation method based on compression domain, to the lens mutation and camera lens in video
Gradual change is detected.On the basis of satisfactory recall ratio and precision ratio is ensured, shot segmentation is quickly realized.
A kind of shot segmentation method based on X264 compression videos, includes the following steps：
Step 1, decoding extraction frame number, frame type and intra prediction are macro frame by frame as unit of the GOP of X264 compression videos
Block ratio.The frame type includes I frames, P frames and B frames；The intra prediction macro block ratio refers in a frame image using in frame
The ratio of the macroblock number and all number of macroblocks in the frame image of prediction mode coding.
Step 2, Shot Detection is carried out to I frames, judges whether current I frames occur lens mutation according to the length of GOP to be measured.
Step 3, using the intra prediction macro block ratio of all P frames in the GOP to be measured as sample data, according to sample data
Degree of fluctuation judge whether to need shot segmentation.
Step 4, shot segmentation is carried out based on predicted macroblock ratio.
Further, the step 2 judges that the method whether current I frames occur lens mutation is as follows：
The serial number difference of two neighboring I frames is asked to obtain the length of GOP to be measured.By the GOP length and the maximum length of GOP
Keyint is compared.If the length is less than keyint, judge that lens mutation occurs for current I frames.
Further, the step 3 judges whether to need the method for shot segmentation such as according to the degree of fluctuation of sample data
Under：
The standard deviation of the sample data is obtained.If the standard deviation is less than the threshold value of setting, illustrate the wave of sample data
Traverse degree is smaller, represents that the picture variation in the GOP is smaller, there is no camera lenses in the GOP to convert, and does not need to camera lens point
It cuts；Otherwise, illustrate that the degree of fluctuation of sample data is larger, represent that the picture in the GOP changes greatly, exist in the GOP
Camera lens is converted, and needs to carry out shot segmentation.
Further, method of the step 4 based on predicted macroblock ratio progress shot segmentation is as follows：
Lens mutation threshold value T1 and the threshold value T2 of gradual shot are set；
Frame in the GOP is divided into suspicion frame and normal frame：P frame of the intra prediction macro block ratio more than T2 is suspicion
Frame；P frame of the intra prediction macro block ratio less than T2 and B frames are normal frame.At most allow continued presence 2 between every 2 suspicion frames
A normal frame.All frames in GOP to be measured are traversed, are divided into several segments.The segment is divided into two classes：One kind be without
The segment of normal frame, another kind of is the segment containing normal frame；
The threshold value Tw of detection gradual shot is set respectively to the length of the two classes different fragments.By fragment length and accordingly
Threshold value Tw be compared, if fragment length be more than Tw, judge the segment be a gradual shot；Otherwise, if described
The intra prediction macro block ratio of P frames is more than T1 in section, then judges that lens mutation has occurred in the P frames.
Further, the setting method of the lens mutation threshold value T1 and the threshold value T2 of gradual shot are：If sample
The desired value of data adds 0.5 to be more than 0.7, then T1 is set as 0.7；Otherwise, T1 is set as 0.6.Values of the T2 equal to T1 subtracts
0.2。
Further, the setting method of the threshold value Tw is：Segment without normal frame, Tw are set as 5 frames；Containing just
The segment of normal frame, Tw are set as 10 frames.
Compared with prior art, the present invention has the following advantages：
It is proposed by the present invention to be independent from each other for I frames and the detection method of P frames when carrying out shot cut detection,
It is independent of each other to mutual testing result.
The method of the invention carries out cycle detection as unit of GOP, and GOP correlated samples data are counted before detection
Analysis changes picture smaller GOP without shot segmentation, reduces operand.
The method of the invention can carry out shot segmentation to GOP length and the unfixed compression video of structure.Especially
When shot segmentation is happened at I frames, the prior art is generally examined using the variation of predicted macroblock ratio in I frame consecutive frames
It surveys, this requires video has fixed GOP length and structure.And the method for the invention is not then by GOP length and structure
The shot transition detection to I frames only can be realized in limitation using the I frames parameters of itself.Experiment shows using side of the present invention
Method carries out shot segmentation to compression video, has good recall ratio and precision ratio, recall level average and precision ratio can reach respectively
To 96.1% and 98.6%.
Description of the drawings
Fig. 1 is the principle schematic for carrying out shot segmentation based on macro block ratio in the prior art；
Fig. 2 is the method flow diagram that shot segmentation is carried out to I frames；
Fig. 3 is the method flow diagram of embodiment.
Specific embodiment
The specific embodiment of the present invention is described in further detail with reference to the accompanying drawings and examples.Following embodiment
It is served only for illustrating the present invention, but be not limited to the scope of the present invention.
It is a kind of based on X264 compression video shot segmentation method, mutation conversion and gradual change to camera lens in video convert into
Row detection.Two aspects are divided into for the detection of abrupt shot：The detection of I frames is happened to mutation and P is happened to mutation
The detection of frame.If mutation is happened at I frames, flow is divided according to the GOP of X264, is detected by comparing the length of GOP；It is if prominent
Change is happened at P frames, then needs whether the intra prediction macro block ratio P_intra for investigating P frames exceeds the threshold value of lens mutation.And to gradually
Become the threshold value that the detection of camera lens then needs the P_intra for investigating P frames whether to exceed gradual shot.The present invention is extracted as unit of GOP
Code stream information carries out cycle detection.
As shown in figure 3, in specific implementation, it is sequentially completed following steps in a computer：
The first step：Using FFMPEG as experiment porch, input X264 compresses video as cycle tests.Initialize code stream letter
The memory space vector of breath and maximum GOP length keyint.
Second step：The letters such as extraction frame number, frame type and macro-block skip mode during a GOP is decoded frame by frame
It ceases and calculates the macro block ratio P_intra using intra prediction.It will be in the storage to vector of these code stream informations.
Third walks：When decoding is to next I frames, represent that a GOP terminates.Pause decoding.Camera lens is carried out to previous GOP
Segmentation.Step is as follows：
1) the I frame numbers of the I frame numbers of GOP to be measured and previous GOP are done into difference, so as to obtain the length of previous GOP
Degree.This GOP length is compared with keyint.If GOP length is less than keyint, it can determine that the I frames of GOP to be measured have occurred
Lens mutation；Otherwise, with the value of GOP length update keyint.
2) P_intra of all P frames as sample data, calculates expectation and standard deviation using in GOP to be measured.If standard deviation is small
In 0.1, represent that sample data fluctuation is smaller, GOP image contents variation is smaller, judges that no camera lens conversion occurs, skips to the 4th
Step.If standard deviation is greater than or equal to 0.1, represent that sample data fluctuation is larger, which changes greatly, and there are camera lenses
Conversion, needs to be detected.
3) each threshold value of shot segmentation is set.For the threshold value T1 of lens mutation, handled using the method for segmentation.By sample
The desired value of notebook data adds 0.5 imparting T1.If T1 is more than 0.7, T1 is set as 0.7；If T1 is less than 0.7, T1 is set
It is 0.6.For the threshold value T2 of gradual shot, the value of T1 is directly subtracted 0.2 and is obtained.
4) frame in GOP to be measured is divided into two classes：One kind is suspicion frame, i.e. P_intra is more than the P frames of T2；It is another kind of to be
Normal frame, i.e. P_intra are less than the P frames of T2 and B frames.At most allow continued presence 2 normal between regulation each two suspicion frame
Frame.All frames in GOP to be measured are traversed, are divided into several segments being made of continuous suspicion frame, and calculate piece segment length
Degree.
5) fragment length with corresponding threshold value Tw is compared, if fragment length is more than corresponding threshold value Tw, judged
The segment is a gradual shot.It needs exist for segment being divided into two classes：One kind is the segment without normal frame, length threshold
Tw is set as 5 frames；Another kind of is the segment containing normal frame, and length threshold Tw is set as 10.
6) it is less than the segment of respective threshold for length, then needs to carry out the detection of lens mutation.If the wherein P_ of P frames
Intra is more than T1, then judges that lens mutation has occurred in the P frames.
4th step：The variables such as vector, T1, T2 are emptied, skip to second step, all frames until having detected video to be measured.
In order to examine the validity of the method for the invention, tested using the X264 compression videos of 4 groups of different contents.
The effect of shot segmentation algorithm is evaluated usually using recall ratio (Recall) and precision ratio (Precision) the two indexs.
Test result is as shown in table 1.As shown in Table 1, the recall ratio of the 1st cycle tests and precision ratio have reached 100%, 4 groups of sequences
Recall level average and precision ratio be respectively 96.1% and 98.6%.
The recall ratio and precision ratio of 1 cycle tests of table
| Cycle tests | Recall ratio | Precision ratio |
| 1 | 100% | 100% |
| 2 | 98.1% | 96.2% |
| 3 | 93.2% | 98.2% |
| 4 | 93.2% | 100% |
| Mean value | 96.1% | 98.6% |
This example demonstrates that：Method proposed by the present invention can compress video to X264 and carry out effective shot segmentation, energy
Enough obtain preferable recall ratio and precision ratio.Shot segmentation algorithm based on compression domain is upper at runtime to be relatively based on uncompressed domain
Algorithm also have some superiority.
Claims (3)
1. a kind of shot segmentation method based on X264 compression videos, it is characterised in that include the following steps：
Step 1, frame number, frame type and intra prediction are extracted in decoding frame by frame as unit of the image group GOP of X264 compression videos
Macro block ratio；
The frame type includes I frames, P frames and B frames；The intra prediction macro block ratio refers to use intra prediction in a frame image
The ratio of the macroblock number of pattern-coding and all macroblock numbers in the frame image；
Step 2, Shot Detection is carried out to the I frames in GOP to be measured, judges whether current I frames occur mirror according to the length of GOP to be measured
Head process becomes；
Step 3, using the intra prediction macro block ratio P_intra of all P frames in the GOP to be measured as sample data, according to sample
The degree of fluctuation of data judges whether to need shot segmentation；
Step 4, shot segmentation is carried out based on predicted macroblock ratio；
The step 2 judges that the method whether current I frames occur lens mutation is as follows：
The serial number difference of two neighboring I frames is asked to obtain the length of GOP to be measured；By the GOP length and the maximum length of GOP
Keyint is compared；If the length is less than keyint, judge that lens mutation occurs for current I frames；
The step 3 judges whether to need the method for shot segmentation as follows according to the degree of fluctuation of sample data：
The standard deviation of the sample data is obtained；If the standard deviation is less than the threshold value of setting, illustrate the fluctuation journey of sample data
Degree is smaller, represents that the picture variation in the GOP is smaller, there is no camera lenses in the GOP to convert, and does not need to shot segmentation；It is no
Then, illustrate that the degree of fluctuation of sample data is larger, represent that the picture in the GOP changes greatly, there are camera lenses in the GOP to turn
It changes, needs to carry out shot segmentation；
The method that the step 4 carries out shot segmentation based on predicted macroblock ratio is as follows：
The lens mutation threshold value T1 of shot segmentation and the threshold value T2 of gradual shot are set；
Frame in the GOP is divided into suspicion frame and normal frame：P frames of the P_intra more than T2 is suspicion frame；P_intra is less than
The P frames and B frames of T2 is normal frame；At most allow 2 normal frames of continued presence between every 2 suspicion frames；It traverses in GOP to be measured
All frames are divided into several segments；The segment is divided into two classes：One kind for the segment without normal frame, it is another kind of be containing
There is the segment of normal frame；
The threshold value Tw of detection gradual shot is set respectively to the length of the two classes different fragments；By fragment length and corresponding threshold
Value Tw is compared, if fragment length is more than corresponding threshold value Tw, it is a gradual shot to judge the segment；Otherwise, if
The P_intra of P frames is more than T1 in the segment, then judges that lens mutation has occurred in the P frames.
2. the shot segmentation method according to claim 1 based on X264 compression videos, which is characterized in that the camera lens is dashed forward
Become the setting method of the threshold value T2 of threshold value T1 and gradual shot into：If the desired value of sample data is more than 0.7 after adding 0.5,
T1 is then set as 0.7；Otherwise, T1 is set as 0.6；Values of the T2 equal to T1 subtracts 0.2.
3. the shot segmentation method according to claim 1 based on X264 compression videos, which is characterized in that the threshold value Tw
Setting method be：Segment without normal frame, threshold value Tw are set as 5 frames；Segment containing normal frame, threshold value Tw are set as 10
Frame.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510253354.1A CN104869403B (en) | 2015-05-18 | 2015-05-18 | A kind of shot segmentation method based on X264 compression videos |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510253354.1A CN104869403B (en) | 2015-05-18 | 2015-05-18 | A kind of shot segmentation method based on X264 compression videos |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN104869403A CN104869403A (en) | 2015-08-26 |
| CN104869403B true CN104869403B (en) | 2018-06-12 |
Family
ID=53914862
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201510253354.1A Expired - Fee Related CN104869403B (en) | 2015-05-18 | 2015-05-18 | A kind of shot segmentation method based on X264 compression videos |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN104869403B (en) |
Families Citing this family (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106231348B (en) * | 2016-09-19 | 2019-06-11 | 浙江宇视科技有限公司 | A kind of back method of GOP data, device and system |
| CN107437076B (en) * | 2017-08-02 | 2019-08-20 | 逄泽沐风 | The method and system that scape based on video analysis does not divide |
| CN109561310B (en) * | 2017-09-26 | 2022-09-16 | 腾讯科技（深圳）有限公司 | Video coding processing method, device, equipment and storage medium |
| CN110119735A (en) * | 2018-02-06 | 2019-08-13 | 上海全土豆文化传播有限公司 | The character detecting method and device of video |
| CN111327945B (en) | 2018-12-14 | 2021-03-30 | 北京沃东天骏信息技术有限公司 | Method and apparatus for segmenting video |
| CN110677657B (en) * | 2019-11-01 | 2022-02-18 | 杭州当虹科技股份有限公司 | Scene segmentation method for content adaptive coding |
| CN111966859A (en) * | 2020-08-27 | 2020-11-20 | 司马大大（北京）智能系统有限公司 | Video data processing method and device and readable storage medium |
| CN112714336B (en) * | 2020-12-01 | 2022-12-02 | 万兴科技集团股份有限公司 | Video segmentation method and device, electronic equipment and computer readable storage medium |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101184226A (en) * | 2006-11-14 | 2008-05-21 | 腾讯科技（深圳）有限公司 | Method and device of inserting video key frame |
| CN102630013A (en) * | 2012-04-01 | 2012-08-08 | 北京捷成世纪科技股份有限公司 | Bit rate control video compression method and device on basis of scene switching |
| CN103826121A (en) * | 2013-12-20 | 2014-05-28 | 电子科技大学 | Scene switching detection based code rate control method in low-delay video coding |
-
2015
- 2015-05-18 CN CN201510253354.1A patent/CN104869403B/en not_active Expired - Fee Related
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101184226A (en) * | 2006-11-14 | 2008-05-21 | 腾讯科技（深圳）有限公司 | Method and device of inserting video key frame |
| CN102630013A (en) * | 2012-04-01 | 2012-08-08 | 北京捷成世纪科技股份有限公司 | Bit rate control video compression method and device on basis of scene switching |
| CN103826121A (en) * | 2013-12-20 | 2014-05-28 | 电子科技大学 | Scene switching detection based code rate control method in low-delay video coding |
Non-Patent Citations (2)
| Title |
|---|
| 《基于H.264/AVC压缩域的实时视频镜头分割算法》;洪夏俊等;《电脑知识与技术》;20090228;第5卷(第4期);944-945 * |
| 《视频检索常用的镜头分割方法的研究》;朱耀麟等;《视频应用与工程》;20140331;第38卷(第3期);178-181 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN104869403A (en) | 2015-08-26 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN104869403B (en) | A kind of shot segmentation method based on X264 compression videos | |
| US6618507B1 (en) | Methods of feature extraction of video sequences | |
| EP1021042B1 (en) | Methods of scene change detection and fade detection for indexing of video sequences | |
| US8345750B2 (en) | Scene change detection | |
| CN101453649B (en) | Key frame extracting method for compression domain video stream | |
| JPH11252509A (en) | Detector for cut point of moving image | |
| US9883200B2 (en) | Method of acquiring neighboring disparity vectors for multi-texture and multi-depth video | |
| US20080095232A1 (en) | Method and apparatus for implementing b-picture scene changes | |
| CN101978698A (en) | Method and apparatus for encoding and decoding image | |
| JP2000217121A (en) | Method for detecting scene change by processing video data in compressed form for digital image display | |
| Dey et al. | Robust background subtraction for network surveillance in H. 264 streaming video | |
| GB2459671A (en) | Scene Change Detection For Use With Bit-Rate Control Of A Video Compression System | |
| CN109194955B (en) | Scene switching detection method and system | |
| KR101149522B1 (en) | Apparatus and method for detecting scene change | |
| Peixoto et al. | Fast H. 264/AVC to HEVC transcoding based on machine learning | |
| Duarte et al. | Fast affine motion estimation for vvc using machine-learning-based early search termination | |
| Laumer et al. | Moving object detection in the H. 264/AVC compressed domain | |
| CN113422959A (en) | Video encoding and decoding method and device, electronic equipment and storage medium | |
| CN100373952C (en) | Rapid kinematics estimation of video frequency object based on MPEG-4 | |
| CN104168482A (en) | Method and device for video coding and decoding | |
| Jie et al. | A novel scene change detection algorithm for H. 264/AVC bitstreams | |
| Anegekuh et al. | Encoded bitstream based video content type definition for HEVC video quality prediction | |
| Huangyuan et al. | Learning based fast H. 264 to H. 265 transcoding | |
| CN103248889B (en) | A kind of image processing method and device | |
| JP3571200B2 (en) | Moving image data cut detection apparatus and method, and recording medium |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| EXSB | Decision made by sipo to initiate substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| CF01 | Termination of patent right due to non-payment of annual fee | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20180612 Termination date: 20190518 |