CN104702956B - A kind of background modeling method towards Video coding - Google Patents
A kind of background modeling method towards Video coding Download PDFInfo
- Publication number
- CN104702956B CN104702956B CN201510129779.1A CN201510129779A CN104702956B CN 104702956 B CN104702956 B CN 104702956B CN 201510129779 A CN201510129779 A CN 201510129779A CN 104702956 B CN104702956 B CN 104702956B
- Authority
- CN
- China
- Prior art keywords
- background
- block
- image block
- image
- coding
- 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
- 238000000034 method Methods 0.000 title claims abstract description 29
- 238000013139 quantization Methods 0.000 claims description 6
- 230000008569 process Effects 0.000 claims description 3
- 230000006835 compression Effects 0.000 abstract description 5
- 238000007906 compression Methods 0.000 abstract description 5
- 238000004364 calculation method Methods 0.000 description 3
- 241001269238 Data Species 0.000 description 2
- 239000002699 waste material Substances 0.000 description 2
- 230000003044 adaptive effect Effects 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 235000013399 edible fruits Nutrition 0.000 description 1
- 230000006870 function Effects 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 230000007774 longterm Effects 0.000 description 1
- 239000000463 material Substances 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 230000000737 periodic effect Effects 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 238000005303 weighing Methods 0.000 description 1
Abstract
The invention discloses a kind of background modeling method towards Video coding, by optionally being carried out on the basis of image block, local background's modeling, local background are updated, local multi background is kept, be difficult to determine by the best opportunity for overcoming context update present in the existing background modeling method in units of frame, under particular case background modeling it is difficult, to periodicity background modeling inefficiency the problems such as, so as to reduce the expense of background modeling, improve the forecast quality of background model, the final overall compression efficiency for improving Video coding.
Description
Technical field
The invention belongs to technical field of video coding, more particularly to a kind of new background modeling side based on image block
Method.
Background technology
As the extensive use of multimedia video technique is, it is necessary to which the video data volume for storing or transmitting is also increasing.For
Storage or transmission cost are saved, it is necessary to these video datas of Efficient Compression.
At present, video encoding standard is applied and designed primarily directed to generic video, not for fixing camera
Video Applications(Such as video monitoring, video conference)Make fully optimization.Therefore, can be abundant when encoding these video datas
This feature is fixed using camera, existing coding standard is improved, further improves compression efficiency.
One of improved method is to carry out background modeling.Long term reference frame is used as by the background image that modeling is obtained,
Background redundancy can be effectively eliminated, so as to improve code efficiency.Therefore, Many researchers are drawn in existing video encoding standard
Background modeling method is entered, such as【Document 1】.
But, these methods are while code efficiency is improved, and there is also some areas for improvement, major embodiment
:
1st, the background area blocked for a long time for passive movement prospect, such as crowded stream of people or wagon flow, the fountain persistently erupted
Deng because these background areas are always being moved or are being blocked, therefore hardly resulting in preferable modeling result.But in units of frame
Background modeling method these regions must be modeled to obtain a complete background frames, and to the modeling in these regions
As a result often obscure, also will not be substantially referenced in coding.Therefore, code check is not only wasted to region modeling,
Waste the calculating time.
2nd, due to many reasons, background can also change, thus background modeling for a period of time after, it is necessary to upgrade in time,
New background is modeled to substitute out-of-date background.Due to background content many times be local updating, therefore it is existing with
Frame is that unit progress background modeling may result in a double-barreled question:It is used to represent that the code check of background is too high if frequent updating
(The background area not changed will also be modeled again, and modeling result needs to take code check), reduce overall coding efficiency;Such as
Then forecast quality can also be reduced fruit reduction renewal frequency, equally also reduce overall coding efficiency.
3rd, existing background modeling method synchronization substantially only preserves a background frames(If preserving multiple complete back ofs the body
Scape frame then needs a large amount of memory spaces), thus effectively periodic background can not be modeled(Such as hidden repeatedly by vehicle
The road surface that gear reappears again, generally requires modeling repeatedly, wastes code check)
Therefore, it is necessary to propose a kind of new technical scheme to solve the above problems.
【Document 1】Xianguo Zhang. Background-Modeling-Based Adaptive Prediction
for Surveillance Video Coding. IEEE Transaction on Image Processing. 2014.
The content of the invention
In order to solve above-mentioned technical problem, the present invention proposes a kind of efficient block-based back of the body towards Video coding
Scape modeling method.
The technical solution adopted in the present invention is:A kind of background modeling method towards Video coding, it is characterised in that bag
Include following steps:
Step 1:For the coding of the video sequence from fixing camera, carried out on the level of image block rather than frame
Background modeling;Meanwhile, for the ease of decoding, background modeling is carried out by display order on reconstructed image sequence;Reconstructed image sequence
All images are divided into image block by BxB pixels in row, therefore a video sequence can be regarded as MxN image block sequence,
Wherein MxB=W, NxB=H, W are picture traverse, and H is picture altitude;
Step 2:Each image block sequence carries out using during background modeling, image block sequence background modeling independently of one another
To data structure have powerful connections block judge caching and background block caching;Described background block judges to cache the team for being length for K blocks
Row, described background block caching is the queue that length is L blocks, is all sky when the two is initial;Image block sequence background modeling process
Including following sub-step:
Step 2.1:Current reconstructed image block is added to the tail of the queue that background block judges caching;
Step 2.2:If current reconstructed image block is similar to some background blocks, first delete all existing with currently weighing
The similar background block of structure image block, then current reconstructed image block is added to the tail of the queue of background block buffer queue;This flow is finished;
Otherwise, then the judgement of following step 2.3 is proceeded;
Step 2.3:If background block judges existing K blocks in caching, judge whether they are all similar
If, then it is assumed that current reconstructed image block is background block, first deletes all existing with current reconstructed image block phase
As background block, then current reconstructed image block is added to the tail of the queue of background block buffer queue;This flow is finished;
If it is not, then current reconstructed image block is not background block;
Step 3:The original picture block to be encoded of inter-frame encoding is used for each, the accurate interframe of advanced rower is searched
Each background block in rope, then the background image block caching of additional searching correspondence position;If certain background block is optimal reference
Block, then record its sequence number in background image block caching;All backgrounds actually referred to when encoding a frame original image
Sequence number of the block in respective background image block caching constitutes background block selection bitmap, after the bitmap is compressed by some way,
It is stored in code stream and is transmitted to decoding end, identical background modeling algorithm during with coding is used during decoding, position is selected with reference to background block
Figure, reconstructs identical virtual reference frame, is correctly decoded with realizing;
Step 4:After present frame coding is finished, while also having obtained reconstructed image.The reconstructed image is first square as described in step 1
Formula is divided into image block, and mode carries out background modeling to each image block as described in step 2 again, if background block, then in step
Additional alternative reference block is used as in rapid 3.So repeat until all frames coding is finished.
Preferably, the optimal reference block described in step 3 is the minimum block of Coding cost, have one in existing coding standard
Individual cost function dedicated calculation Coding cost, has a variety of calculations alternative, user can pass through coding profile before encoding
Specify, the present invention continues to use the calculation that user currently selects herein.
Preferably, the bitmap described in step 3 is compressed by some way, uses arithmetic coding mode and carry out
Compression.
Preferably, in step 2 after the completion of two field picture coding, all background blocks for meeting above-mentioned condition add in reconstructed frame
Enter into corresponding background image block caching, can be eliminated if space is overflowed by certain rule.
It can then be eliminated preferably, described space is overflowed by certain rule, use first-in first-out rule.
Preferably, two whether similar determination methods of image block are in step 2:
Step 2.2.1:Two image blocks subtract each other, and obtain residual error;
Step 2.2.2:Enter line translation, quantization to residual error;
Step 2.2.3:Judge whether residual error can be represented after entering line translation, quantization with SKIP modes
If, then it is assumed that two image blocks are similar;
If not, then it is assumed that two image blocks are dissimilar.
A kind of background modeling method that the present invention is provided, by optionally carrying out local background on the basis of image block
Modeling, local background are updated, local multi background is kept, to overcome present in the existing background modeling method in units of frame
Above mentioned problem, so as to reduce the expense of background modeling, improves the forecast quality of background model, the final totality for improving Video coding
Compression efficiency.
Brief description of the drawings
Fig. 1:The flow chart of the embodiment of the present invention;
Fig. 2:The interframe search routine figure of the extension of the embodiment of the present invention.
Embodiment
Understand for the ease of those of ordinary skill in the art and implement the present invention, below in conjunction with the accompanying drawings and embodiment is to this hair
It is bright to be described in further detail, it will be appreciated that implementation example described herein is merely to illustrate and explain the present invention, not
For limiting the present invention.
A kind of background modeling method towards Video coding provided see Fig. 1 and Fig. 2, the present invention, comprises the following steps:
Step 1:For the coding of the video sequence from fixing camera, carried out on the level of image block rather than frame
Background modeling;Meanwhile, for the ease of decoding, background modeling is carried out by display order on reconstructed image sequence;Reconstructed image sequence
All images are divided into image block by BxB pixels in row(Such as the macro block of 16x16 pixels, or other sizes), therefore a video
Sequence can be regarded as MxN image block sequence, wherein MxB=W, and NxB=H, W is picture traverse, and H is picture altitude;
Step 2:Each image block sequence carries out using during background modeling, image block sequence background modeling independently of one another
To data structure have powerful connections block judge caching and background block caching;Described background block judges to cache the team for being length for K blocks
Row, if overflowed, can eliminate head of the queue element;Described background block caching is the queue that length is L blocks, if overflowed, can be eliminated
Head of the queue element;Image block sequence background modeling process includes following sub-step:
Step 2.1:Current image block is added to the tail of the queue that background block judges caching;
Step 2.2:If current image block is similar to some background blocks, the tail of the queue of background block buffer queue is added to,
This flow is finished;
Otherwise, then the judgement of following step 2.3 is proceeded;
Two whether similar determination methods of image block are in step 2.2:
Step 2.2.1:Two image blocks subtract each other, and obtain residual error;
Step 2.2.2:Enter line translation, quantization to residual error;
Step 2.2.3:Judge whether residual error can be represented after entering line translation, quantization with SKIP modes
If, then it is assumed that two image blocks are similar;
If not, then it is assumed that two image blocks are dissimilar;
Step 2.3:If background block judges existing K blocks in caching, judge whether they are all similar
If, then it is assumed that current image block is background block, first deletes all existing backgrounds similar to current image block
Block, then current image block is added to the tail of the queue of background block buffer queue;This flow is finished;
If it is not, not being then background block;
Step 3:The original picture block to be encoded of inter-frame encoding is used for each, the accurate interframe of advanced rower is searched
Each background block in rope, then the background image block caching of additional searching correspondence position;If certain background block is optimal reference
Block(That is the minimum block of Coding cost), then its sequence number in background image block caching is recorded;It is all by reality in one two field picture
Sequence number of the background block of reference in respective background image block caching constitutes background block selection bitmap, and the bitmap uses arithmetic coding
After mode is compressed, it is stored in code stream and is transmitted to decoding end, identical background modeling algorithm during with coding, knot is used during decoding
Background block selection bitmap is closed, identical virtual reference frame is reconstructed, is correctly decoded with realizing;
Step 4:After present frame coding is finished, while also having obtained reconstructed image;The reconstructed image is first square as described in step 1
Formula is divided into image block, and mode carries out background modeling to each image block as described in step 2 again, if background block, then in step
Additional alternative reference block is used as in rapid 3;So repeat until all frames coding is finished.
After the completion of the one two field picture coding of the present embodiment, all background blocks for meeting above-mentioned condition are added to phase in reconstructed frame
, can be by certain rule if space is overflowed in the background image block caching answered(Such as first in first out)Eliminate.
Can be that background modeling is carried out on the level of image block per two field picture by above-mentioned steps.On the one hand, for every
For two field picture, according to its picture material, the result of modeling is not necessarily MxN background block, but is probably less than MxN;
For a modeling period, the modeling result of some tile location is also not necessarily 1, and is probably 0 to arrive
L.
By this method, it is possible to which realization optionally carries out local background's modeling, part on the basis of image block
Context update, local multi background are kept, so as to effectively solve three problems that background technology is mentioned.
It should be appreciated that the part that this specification is not elaborated belongs to prior art.
It should be appreciated that the above-mentioned description for preferred embodiment is more detailed, therefore it can not be considered to this
The limitation of invention patent protection scope, one of ordinary skill in the art is not departing from power of the present invention under the enlightenment of the present invention
Profit is required under protected ambit, can also be made replacement or be deformed, each fall within protection scope of the present invention, this hair
It is bright scope is claimed to be determined by the appended claims.
Claims (6)
1. a kind of background modeling method towards Video coding, it is characterised in that comprise the following steps:
Step 1:For the coding of the video sequence from fixing camera, background is carried out on the level of image block rather than frame
Modeling;Meanwhile, for the ease of decoding, background modeling is carried out by display order on reconstructed image sequence;In reconstructed image sequence
All images are divided into image block by BxB pixels, therefore a video sequence can be regarded as MxN image block sequence, wherein
MxB=W, NxB=H, W are picture traverse, and H is picture altitude;
Step 2:Each image block sequence carries out what is used during background modeling, image block sequence background modeling independently of one another
Data structure have powerful connections block judge caching and background block caching;Described background block judges that caching is length for the queue of K blocks, institute
The background block caching stated is the queue that length is L blocks, is all sky when the two is initial;Image block sequence background modeling process include with
Lower sub-step:
Step 2.1:Current reconstructed image block is added to the tail of the queue that background block judges caching;
Step 2.2:If current reconstructed image block is similar to some background blocks, first delete all existing with current reconstruct image
As the similar background block of block, then current reconstructed image block is added to the tail of the queue of background block buffer queue;This flow is finished;
Otherwise, then the judgement of following step 2.3 is proceeded;
Step 2.3:If background block judges existing K blocks in caching, judge whether they are all similar;
If, then it is assumed that current reconstructed image block is background block, first deletes all existing similar to current reconstructed image block
Background block, then current reconstructed image block is added to the tail of the queue of background block buffer queue;This flow is finished;
If it is not, then current reconstructed image block is not background block;
Step 3:The original picture block to be encoded of inter-frame encoding is used for each, the accurate interframe of advanced rower is searched for,
Each background block during the background image block of additional searching correspondence position is cached again;If certain background block is optimal reference block,
Then record its sequence number in background image block caching;When encoding a frame original image, all background blocks actually referred to exist
Sequence number in respective background image block caching constitutes background block selection bitmap, after the bitmap is compressed, is stored in code stream and passes
Identical background modeling algorithm during with coding is used to decoding end, during decoding, bitmap is selected with reference to background block, reconstructs identical
Virtual reference frame, is correctly decoded with realizing;
Step 4:After present frame coding is finished, while also having obtained reconstructed image;The reconstructed image first draw as described in step 1 by mode
It is divided into image block, mode carries out background modeling to each image block as described in step 2 again, if background block, then in step 3
It is used as additional alternative reference block;So repeat until all frames coding is finished.
2. the background modeling method according to claim 1 towards Video coding, it is characterised in that:Described in step 3
Optimal reference block is the minimum block of Coding cost.
3. the background modeling method according to claim 1 towards Video coding, it is characterised in that:Described in step 3
Bitmap is compressed, and is used arithmetic coding mode and is compressed.
4. the background modeling method according to claim 1 towards Video coding, it is characterised in that:A frame figure in step 2
After the completion of coding, all background blocks for meeting above-mentioned condition are added in corresponding background image block caching in reconstructed frame, if
Space, which is overflowed, then eliminates.
5. the background modeling method according to claim 4 towards Video coding, it is characterised in that:Overflow in described space
Then eliminate, use first-in first-out rule.
6. the background modeling method according to claim 1 towards Video coding, it is characterised in that:Two figures in step 2
As the whether similar determination methods of block are:
Step 2.2.1:Two image blocks subtract each other, and obtain residual error;
Step 2.2.2:Enter line translation, quantization to residual error;
Step 2.2.3:Judge whether residual error can be represented after entering line translation, quantization with SKIP modes;
If, then it is assumed that two image blocks are similar;
If not, then it is assumed that two image blocks are dissimilar.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201510129779.1A CN104702956B (en) | 2015-03-24 | 2015-03-24 | A kind of background modeling method towards Video coding |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201510129779.1A CN104702956B (en) | 2015-03-24 | 2015-03-24 | A kind of background modeling method towards Video coding |
Publications (2)
Publication Number | Publication Date |
---|---|
CN104702956A CN104702956A (en) | 2015-06-10 |
CN104702956B true CN104702956B (en) | 2017-07-11 |
Family
ID=53349668
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201510129779.1A Expired - Fee Related CN104702956B (en) | 2015-03-24 | 2015-03-24 | A kind of background modeling method towards Video coding |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN104702956B (en) |
Families Citing this family (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN105872556B (en) * | 2016-04-11 | 2020-01-03 | 华为技术有限公司 | Video encoding method and apparatus |
CN106791829B (en) * | 2016-11-18 | 2020-01-21 | 华为技术有限公司 | Method and equipment for establishing virtual reference frame |
Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2009032852A1 (en) * | 2007-09-04 | 2009-03-12 | Objectvideo, Inc. | Background modeling with feature blocks |
CN101533515A (en) * | 2009-04-13 | 2009-09-16 | 浙江大学 | Background modeling method based on block facing video monitoring |
CN101833760A (en) * | 2010-03-29 | 2010-09-15 | 中山大学 | Background modeling method and device based on image blocks |
CN101883284A (en) * | 2010-06-21 | 2010-11-10 | 北京大学 | Video encoding/decoding method and system based on background modeling and optional differential mode |
CN101883209A (en) * | 2010-05-31 | 2010-11-10 | 中山大学 | Method by integrating background model and three-frame difference to detect video background |
CN102006473A (en) * | 2010-11-18 | 2011-04-06 | 无锡中星微电子有限公司 | Video encoder and encoding method, and video decoder and decoding method |
-
2015
- 2015-03-24 CN CN201510129779.1A patent/CN104702956B/en not_active Expired - Fee Related
Patent Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2009032852A1 (en) * | 2007-09-04 | 2009-03-12 | Objectvideo, Inc. | Background modeling with feature blocks |
CN101533515A (en) * | 2009-04-13 | 2009-09-16 | 浙江大学 | Background modeling method based on block facing video monitoring |
CN101833760A (en) * | 2010-03-29 | 2010-09-15 | 中山大学 | Background modeling method and device based on image blocks |
CN101883209A (en) * | 2010-05-31 | 2010-11-10 | 中山大学 | Method by integrating background model and three-frame difference to detect video background |
CN101883284A (en) * | 2010-06-21 | 2010-11-10 | 北京大学 | Video encoding/decoding method and system based on background modeling and optional differential mode |
CN102006473A (en) * | 2010-11-18 | 2011-04-06 | 无锡中星微电子有限公司 | Video encoder and encoding method, and video decoder and decoding method |
Also Published As
Publication number | Publication date |
---|---|
CN104702956A (en) | 2015-06-10 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN101087414B (en) | Image coding apparatus and image coding method | |
CN102547296B (en) | Motion estimation accelerating circuit and motion estimation method as well as loop filtering accelerating circuit | |
CN104967850B (en) | The method and apparatus that image is coded and decoded by using big converter unit | |
CN106254868B (en) | Code rate controlling method for video coding, apparatus and system | |
CN103141092B (en) | The method and apparatus carrying out encoded video signal for the super-resolution based on example of video compress use motion compensation | |
CN104469362A (en) | Method and apparatus for encoding and decoding motion vector | |
CN102217314A (en) | Methods and apparatus for video imaging pruning | |
CN107018418A (en) | Reference data method for reusing, band width estimation method and associated video decoder | |
US20120230594A1 (en) | Adaptive picture rotation | |
CN101883284B (en) | Video encoding/decoding method and system based on background modeling and optional differential mode | |
JP2009521874A (en) | Geometric image representation and compression | |
CN104702956B (en) | A kind of background modeling method towards Video coding | |
CN103596010B (en) | Video coding and decoding system based on dictionary learning and compressed sensing | |
CN109429064A (en) | A kind of method and apparatus handling video data | |
CN106688234A (en) | Scalable transform hardware architecture with improved transpose buffer | |
CN113132727B (en) | Scalable machine vision coding method and training method of motion-guided image generation network | |
CN104702959B (en) | A kind of intra-frame prediction method and system of Video coding | |
CN102148990A (en) | Device and method for predicting motion vector | |
CN113709504B (en) | Image processing method, intelligent terminal and readable storage medium | |
CN102196269B (en) | Grayscale image sequence coding method for traffic access monitoring system | |
CN113068041B (en) | Intelligent affine motion compensation coding method | |
CN102843555A (en) | Intra-frame prediction method and system | |
CN104363454A (en) | Method and system for video coding and decoding of high-bit-rate images | |
CN111901592B (en) | Weight updating method and device for predictive coding neural network | |
CN105635731B (en) | The inter-frame predicated reference point preprocess method of efficient video coding |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into 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: 20170711 |