CN102496165A - Method for comprehensively processing video based on motion detection and feature extraction - Google Patents
Method for comprehensively processing video based on motion detection and feature extraction Download PDFInfo
- Publication number
- CN102496165A CN102496165A CN2011104025767A CN201110402576A CN102496165A CN 102496165 A CN102496165 A CN 102496165A CN 2011104025767 A CN2011104025767 A CN 2011104025767A CN 201110402576 A CN201110402576 A CN 201110402576A CN 102496165 A CN102496165 A CN 102496165A
- Authority
- CN
- China
- Prior art keywords
- current block
- block
- feature extraction
- motion detection
- frame
- 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.)
- Pending
Links
Abstract
The invention discloses a method for comprehensively processing a video based on motion detection and feature extraction and belongs to the technical field of image processing. By the motion detection and the feature extraction, extracted image feature information is applied to a plurality of subsequent modules, so that the quality of an image can be effectively improved, occupation of resources of a processor is reduced, the real-time performance of computation is improved, hardware cost and system complexity are reduced, and comprehensive integrated processing service can be realized on the same processor.
Description
Technical field
The present invention relates to technical field of image processing, especially relate to a kind of comprehensive method for processing video frequency based on motion detection and feature extraction.
Background technology
The digital video coding system is transmitted four parts and is constituted by: video acquisition, video pre-treatment, video coding, code stream usually.Wherein video acquisition, video pre-treatment, video coding all can produce significant impact to video quality.
In video system, can also the integrated video intelligent analysis module.The vision signal of this module analysis input, output can be by the information that the people discerned and accepted again.Intelligent video analysis commonly used is used has mobile detection, car plate identification, recognition of face, circumference detection etc.
Analog video signal can cause the motion parts of image the edge sawtooth to occur owing to adopted horizontal-interlace technique, has a strong impact on picture quality, can remove the interlacing effect through modes such as motion compensation.Common way is that the stationary part maintenance is motionless, carries out in the frame or interframe interpolation to motion parts, thereby can effectively keep the image detail of stationary part, removes the edge sawtooth of motion parts simultaneously.
Noise also can produce picture quality and have a strong impact on, and therefore need carry out denoising so that good video source is provided for scrambler.Can adopt in the frame or the mode of interframe is removed to noise.The data in the same frame are only used in denoising in the frame, remove noise through going modes such as high frequency.The interframe denoising then combines processing with adjacent series of frames, thereby can obtain better denoising effect.During practical operation, can dual mode be combined, promptly the motion parts of image done denoising in the frame, stationary part is done the interframe denoising.
In video coding link, scrambler is through motion search and motion compensation generation reference picture, produces residual image after reference picture and current images acquired are done difference, again to residual coding.And motion search and compensation tache can take a large amount of processor resources usually, can reach more than 50% of wastage in bulk or weight of coding.
In current processing system for video, but integrated intelligent analysis module also.Intellectual analysis commonly used is used to be had: mobile detection, car plate location and identification, the detection of people's face, flow detection etc.Intellectual analysis can be widely used in fields such as intelligent transportation, retail shop are special, storehouse safety, prison.
In processing system for video in the past, above various piece normally independently constitutes a big disposal system with each module series connection.But because data volume is very big in the processing system for video, calculated amount is also very big, and a such system can consume a large amount of processor resources, influences real-time, and causes system to become more complicated.
Summary of the invention
In order to overcome the above-mentioned shortcoming of prior art; The invention provides a kind of comprehensive method for processing video frequency based on motion detection and feature extraction; Feature extraction and motion detection block as utility module, are extracted the characteristics such as texture, color, gradient, statistics of image, and combine motion detection; Motion detection and feature extraction result are applied in interlacing, denoising, video coding, the intellectual analysis; Thereby can effectively reduce operand, improve system real time, and reduce system complexity and hardware cost.
The technical solution adopted for the present invention to solve the technical problems is: a kind of comprehensive method for processing video frequency based on motion detection and feature extraction comprises the steps:
The first step, read a piece of current video frame;
Second goes on foot, extracts characteristics of image commonly used, and writes down one or multinomial characteristic according to the needs of subsequent applications, and these characteristics supply each subsequent module public;
The 3rd step, through to second step obtain one or multinomial characteristic choose with relatively, judge whether current block is moving mass, and it is public that judged result is offered subsequent module as input;
The 4th step, go interlacing: if current block is a static block, then current block keeps motionless; If current block is a moving mass, then remove the interlacing effect through the frame interpolate value;
The 5th step, denoising:, then will go piece and reference block after the interlacing to do the interframe denoising if current block is a static block; If current block is a moving mass, then do denoising in the frame;
The 6th step, coding: in coder side, if current block is a static block, then getting motion vector is 0, gets residual error then and encodes; If current block is a moving mass, then handle according to amount of exercise size or other common feature;
The 7th step, intellectual analysis;
The 8th step, the repetition first step are until handling current video frame.
Compared with prior art; Good effect of the present invention is: through a public characteristic extracting module, the image feature information that extracts is applied in follow-up a plurality of modules, thereby can effectively improves picture quality; Reducing processor resource takies; Improve the real-time of computing, reduce hardware cost and system complexity, help on same processor, realizing that comprehensive integral type manages business.
Embodiment
Patent of the present invention relates in the process of video acquisition, coding, intellectual analysis, digital video signal is gone multiple processing and application such as interlacing, denoising, coding, intellectual analysis.
A kind of comprehensive method for processing video frequency based on motion detection and feature extraction comprises the steps:
The first step, read a blocks of data of current video frame, block size is not strict with, to SD and get 4x4 or 8x8 with the piece of hypograph; The piece of high-definition image is got 4x4 or 8x8 or 16x16;
Second goes on foot, common feature is extracted:
Common feature comprises that current block and reference block get differentiated SAD (summary of absolute difference; That is, absolute error with), the color characteristic of current block, the graded characteristic of current block, the edge complexity characteristics of current block, the statistical nature of current block etc.Wherein, the piece on desirable previous output frame of reference block or the background frames on the same position, background frames can produce through certain algorithm usually, and obtain to upgrade through constantly reading new data.
These characteristics can write down one or multinomial according to the needs of subsequent applications.For example to carry out car plate location, just can write down its color characteristic, possibly have the zone through the color judgment car plate to each piece.To Feature Extraction, be not limited to several kinds mentioned above, related application also is not limited to above several kinds of mentioning.
The 3rd the step, judge whether current block is moving mass:
Through to one or multinomial characteristic choose with relatively, can judge whether this piece is moving mass.For the purpose of meticulous, also can do classification to amount of exercise, for example be divided into large, medium and small three kinds of amounts of exercise, also can be divided into more multistage.The result of motion detection offers interlacing, denoising, coding, intelligent analysis module use as input.
The 4th goes on foot, goes interlacing:
If this piece is a static block, then this data block keeps motionless, thereby can effectively keep image detail.If this piece is a moving mass, then remove the interlacing effect through the frame interpolate value.Can choose a field data during concrete operations, go out another field data with this field data interpolation.
The 5th step, denoising:
Do denoising again to going interlacing data afterwards.
(1) if this piece is a static block, then will go piece and reference block after the interlacing to do the interframe denoising, adoptable mode is the interframe weighted filtering, the filtering formula is following:
P(i,j)?= Pref(i,j)?*?Wref?+?Pcur(i,j)?*?Wcur；
Wherein, (i j) is the reference block pixel value to Pref, and Wref is the reference block weights, and (i j) is current block piece pixel value to Pcur, and Wcur is the current pixel weighted value, and i, j are pixel coordinate.Pcur (i, j) and Pref (i, j) satisfy constraint Pcur (i, j)+Pref (i, j)=1.Usually to static block, (i is 0.8 j) to optional Pcur, elects 0.5 as during little amount of exercise, selects 0.3 during middle amount of exercise, selects 0 during large amount of exercise.
(2), then do denoising in the frame if this piece is a moving mass.Can adopt medium filtering, frequency domain filtering, also can adopt the weighting of sliding window method, be the center with this data block promptly; By the pixel moving window; Window size is identical with this data block, and the data in the window and this blocks of data are done comparison, chooses different weights according to comparative result and carries out weighting.
The 6th step, coding:
In coder side, after judging that this piece is static block, can directly get motion vector is 0, gets residual error then and encodes.If this piece is a moving mass, can carry out different disposal according to the amount of exercise size.To little amount of exercise, middle amount of exercise, can directly near 0 vector, carry out motion search, seek optimal match point.If be large amount of exercise, can directly carry out infra-frame prediction and do not re-use inter prediction.Also can be according to other characteristics, like characteristics such as color, edge textures, search for again after choosing suitable motion search area.
The 7th step, intellectual analysis:
Use for intellectual analysis, can further analyze and use according to feature extraction result to this piece.Specific as follows:
For mobile detection, then construct a binary map picture frame in advance.If this piece is a moving mass, then set on the correspondence position of this frame representes that this position is a moving mass.Like this piece is static block, and then this position is changed to " 0 ", representes that this position is a static block.Block-by-block scanning back forms bianry image like this, this image is handled can be accomplished mobile detection.
When doing car plate identification, then can further discern judgement, do the Primary Location of car plate according to texture, the color characteristic of this piece.
The intellectual analysis of other types is used also can adopt similar processing mode.Promptly block feature is provided, by intelligent analysis module the characteristic of extracting is carried out further analyzing and processing again, and finally export the result by characteristic extracting module.
The 8th the step, get back to the first step, carry out the processing of next piece.So circulation is up to handling a frame.
Claims (6)
1. the comprehensive method for processing video frequency based on motion detection and feature extraction is characterized in that, comprises the steps:
The first step, read a piece of current video frame;
Second goes on foot, extracts characteristics of image commonly used, and writes down one or multinomial characteristic according to the needs of subsequent applications, and these characteristics supply each subsequent module public;
The 3rd step, through to second step obtain one or multinomial characteristic choose with relatively, judge whether current block is moving mass, and it is public that judged result is offered subsequent module as input;
The 4th step, go interlacing: if current block is a static block, then current block keeps motionless; If current block is a moving mass, then remove the interlacing effect through the frame interpolate value;
The 5th step, denoising:, then will go piece and reference block after the interlacing to do the interframe denoising if current block is a static block; If current block is a moving mass, then do denoising in the frame;
The 6th step, coding: in coder side, if current block is a static block, then getting motion vector is 0, gets residual error then and encodes; If current block is a moving mass, then handle according to amount of exercise size or other common feature;
The 7th step, intellectual analysis;
The 8th step, the repetition first step are until handling current video frame.
2. the comprehensive method for processing video frequency based on motion detection and feature extraction according to claim 1 is characterized in that: said characteristics of image commonly used comprises that current block and reference block get the edge complexity characteristics of the graded characteristic of the color characteristic of differentiated SAD, current block, current block, current block, the statistical nature of current block.
3. the comprehensive method for processing video frequency based on motion detection and feature extraction according to claim 2 is characterized in that: the piece on desirable previous output frame of said reference block or the background frames on the same position.
4. the comprehensive method for processing video frequency based on motion detection and feature extraction according to claim 1 is characterized in that: said when judging whether current block is moving mass in the 3rd step, can also carry out classification to amount of exercise.
5. the comprehensive method for processing video frequency based on motion detection and feature extraction according to claim 1 is characterized in that: the method for denoising comprises medium filtering, frequency domain filtering or the weighting of sliding window method in the said frame.
6. the comprehensive method for processing video frequency based on motion detection and feature extraction according to claim 5; It is characterized in that: the weighting of said sliding window method is meant with the current block to be the center; By the pixel moving window; Window size is identical with current block, and the data of data in the window and current block are done comparison, chooses different weights according to comparative result and carries out weighting.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN2011104025767A CN102496165A (en) | 2011-12-07 | 2011-12-07 | Method for comprehensively processing video based on motion detection and feature extraction |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN2011104025767A CN102496165A (en) | 2011-12-07 | 2011-12-07 | Method for comprehensively processing video based on motion detection and feature extraction |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN102496165A true CN102496165A (en) | 2012-06-13 |
Family
ID=46187987
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN2011104025767A Pending CN102496165A (en) | 2011-12-07 | 2011-12-07 | Method for comprehensively processing video based on motion detection and feature extraction |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN102496165A (en) |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103581696A (en) * | 2012-08-10 | 2014-02-12 | 英特尔公司 | Method and system for content-aware multimedia streaming |
| CN103885461A (en) * | 2012-12-21 | 2014-06-25 | 宗经投资股份有限公司 | Movement method for makeup tool of automatic makeup machine |
| CN104253929A (en) * | 2013-06-28 | 2014-12-31 | 广州华多网络科技有限公司 | Video denoising method and video denoising system |
| CN106162196A (en) * | 2016-08-16 | 2016-11-23 | 广东中星电子有限公司 | A kind of video coding framework towards intellectual analysis and method |
| CN108603764A (en) * | 2016-02-17 | 2018-09-28 | 三菱电机株式会社 | Information provider unit, information-providing server and information providing method |
| CN108961186A (en) * | 2018-06-29 | 2018-12-07 | 赵岩 | A kind of old film reparation recasting method based on deep learning |
| CN110502962A (en) * | 2018-05-18 | 2019-11-26 | 翔升（上海）电子技术有限公司 | Mesh object detection method, device, equipment and medium in video flowing |
| CN114401405A (en) * | 2022-01-14 | 2022-04-26 | 安谋科技（中国）有限公司 | Video coding method, medium and electronic equipment |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101146222A (en) * | 2006-09-15 | 2008-03-19 | 中国航空无线电电子研究所 | Motion estimation core of video system |
| CN101321287A (en) * | 2008-07-08 | 2008-12-10 | 浙江大学 | Video encoding method based on movement object detection |
| CN101483746A (en) * | 2008-12-22 | 2009-07-15 | 四川虹微技术有限公司 | Deinterlacing method based on movement detection |
| CN101600061A (en) * | 2009-07-09 | 2009-12-09 | 杭州士兰微电子股份有限公司 | De-interlaced method of video motion-adaptive and device |
| CN101895674A (en) * | 2010-07-02 | 2010-11-24 | 浙江红苹果电子有限公司 | De-interlacing method and device for monitoring video |
| CN101969568A (en) * | 2010-11-16 | 2011-02-09 | 上海大学 | Frame rate up conversion-oriented motion estimation method |
| CN102186046A (en) * | 2011-05-31 | 2011-09-14 | 杭州士兰微电子股份有限公司 | Video image de-interlacing scanning method and device |
-
2011
- 2011-12-07 CN CN2011104025767A patent/CN102496165A/en active Pending
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101146222A (en) * | 2006-09-15 | 2008-03-19 | 中国航空无线电电子研究所 | Motion estimation core of video system |
| CN101321287A (en) * | 2008-07-08 | 2008-12-10 | 浙江大学 | Video encoding method based on movement object detection |
| CN101483746A (en) * | 2008-12-22 | 2009-07-15 | 四川虹微技术有限公司 | Deinterlacing method based on movement detection |
| CN101600061A (en) * | 2009-07-09 | 2009-12-09 | 杭州士兰微电子股份有限公司 | De-interlaced method of video motion-adaptive and device |
| CN101895674A (en) * | 2010-07-02 | 2010-11-24 | 浙江红苹果电子有限公司 | De-interlacing method and device for monitoring video |
| CN101969568A (en) * | 2010-11-16 | 2011-02-09 | 上海大学 | Frame rate up conversion-oriented motion estimation method |
| CN102186046A (en) * | 2011-05-31 | 2011-09-14 | 杭州士兰微电子股份有限公司 | Video image de-interlacing scanning method and device |
Cited By (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103581696A (en) * | 2012-08-10 | 2014-02-12 | 英特尔公司 | Method and system for content-aware multimedia streaming |
| CN103885461A (en) * | 2012-12-21 | 2014-06-25 | 宗经投资股份有限公司 | Movement method for makeup tool of automatic makeup machine |
| CN104253929A (en) * | 2013-06-28 | 2014-12-31 | 广州华多网络科技有限公司 | Video denoising method and video denoising system |
| CN104253929B (en) * | 2013-06-28 | 2019-03-12 | 广州华多网络科技有限公司 | Vedio noise reduction method and its system |
| CN108603764A (en) * | 2016-02-17 | 2018-09-28 | 三菱电机株式会社 | Information provider unit, information-providing server and information providing method |
| CN108603764B (en) * | 2016-02-17 | 2022-08-05 | 三菱电机株式会社 | Information providing device, information providing server, and information providing method |
| CN106162196A (en) * | 2016-08-16 | 2016-11-23 | 广东中星电子有限公司 | A kind of video coding framework towards intellectual analysis and method |
| CN106162196B (en) * | 2016-08-16 | 2020-07-14 | 广东中星电子有限公司 | Intelligent analysis-oriented video coding system and method |
| CN110502962A (en) * | 2018-05-18 | 2019-11-26 | 翔升（上海）电子技术有限公司 | Mesh object detection method, device, equipment and medium in video flowing |
| CN108961186A (en) * | 2018-06-29 | 2018-12-07 | 赵岩 | A kind of old film reparation recasting method based on deep learning |
| CN108961186B (en) * | 2018-06-29 | 2022-02-15 | 福建帝视信息科技有限公司 | Old film repairing and reproducing method based on deep learning |
| CN114401405A (en) * | 2022-01-14 | 2022-04-26 | 安谋科技（中国）有限公司 | Video coding method, medium and electronic equipment |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN102496165A (en) | Method for comprehensively processing video based on motion detection and feature extraction | |
| CN110087087A (en) | VVC interframe encode unit prediction mode shifts to an earlier date decision and block divides and shifts to an earlier date terminating method | |
| CN111355956B (en) | Deep learning-based rate distortion optimization rapid decision system and method in HEVC intra-frame coding | |
| CN102915544B (en) | Video image motion target extracting method based on pattern detection and color segmentation | |
| Yao et al. | Detecting video frame-rate up-conversion based on periodic properties of edge-intensity | |
| CN104573674A (en) | 1D (one-dimensional) barcode recognition for real-time embedded system | |
| CN102752483A (en) | Filtering noise reduction system and filtering noise reduction method based on FPGA (field programmable gate array) platform | |
| CN107623857A (en) | Video coding/decoding method, equipment and generation and the method for stored bits stream | |
| CN102457724B (en) | Image motion detecting system and method | |
| CN105120290A (en) | Fast coding method for depth video | |
| CN101527786A (en) | Method for strengthening definition of sight important zone in network video | |
| Kuanar et al. | Gated fusion network for sao filter and inter frame prediction in versatile video coding | |
| CN104702959B (en) | A kind of intra-frame prediction method and system of Video coding | |
| CN104537685B (en) | One kind carries out automatic passenger flow statisticses analysis method based on video image | |
| CN104396236A (en) | Using depth information to assist motion compensation-based video coding | |
| CN107040786B (en) | A kind of H.265/HEVC video steganalysis method adaptively selected based on space-time characteristic of field | |
| Chen et al. | Moving vehicle detection based on union of three-frame difference | |
| CN102647595B (en) | AVS (Audio Video Standard)-based sub-pixel motion estimation device | |
| CN104243994A (en) | Method for real-time motion sensing of image enhancement | |
| CN106878754B (en) | A kind of 3D video depth image method for choosing frame inner forecast mode | |
| CN101902642B (en) | Quick decision method for H.264 interframe SKIP modes | |
| CN101841722A (en) | Detection device of filtering boundary strength and detection method thereof | |
| CN108024113B (en) | Target ratio self-adaptive compressed domain small target tracking method | |
| CN110472478B (en) | Scene analysis method and system based on optical flow field statistical characteristics | |
| CN103188490A (en) | Combination compensation mode in video coding process |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C12 | Rejection of a patent application after its publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20120613 |