CN103124347B - Vision perception characteristic is utilized to instruct the method for multiple view video coding quantizing process - Google Patents
Vision perception characteristic is utilized to instruct the method for multiple view video coding quantizing process Download PDFInfo
- Publication number
- CN103124347B CN103124347B CN201210402003.9A CN201210402003A CN103124347B CN 103124347 B CN103124347 B CN 103124347B CN 201210402003 A CN201210402003 A CN 201210402003A CN 103124347 B CN103124347 B CN 103124347B
- Authority
- CN
- China
- Prior art keywords
- frame
- viewpoint
- distortion threshold
- follows
- average
- 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 method utilizing vision perception characteristic to instruct coded quantization process.The operating procedure of this method is as follows: (1) reads the brightness value size of each frame of input video sequence, that sets up frequency domain just can distinguish distortion threshold model, (2) prediction of each frame of input video sequence in viewpoint and between viewpoint, (3) discrete cosine transform is carried out to residual error data, (4) quantization step of each macro block in dynamic adjustments present frame, (5) LaGrange parameter in dynamic adjustments rate-distortion optimization process, (6) carry out entropy code to the data quantized, generated code flows through Internet Transmission.The present invention, when ensureing that subjective quality remains unchanged substantially, improves video compression efficiency, is more applicable to transmitting in a network.
Description
Technical field
The present invention relates to multi-view point video encoding and decoding technique field, particularly utilize vision perception characteristic to instruct the method for multiple view video coding quantizing process, be applicable to the encoding and decoding of high definition 3D vision signal.
Background technology
Along with era development, the requirement of people to hearing experience is more and more higher, is not content with existing haplopia two-dimensional video.People are more and more higher for third dimension experience requirements, can experience third dimension, thus expedite the emergence of out the development of multi-vision-point encoding technology from the third dimension of fixed angle to arbitrarily angled.But the data that multiple views requires improve greatly, how effectively to improve video compression efficiency and become study hotspot.At present, video compression technology mainly concentrates on and removes spatial redundancy, time redundancy and statistical redundancy three aspects.Although video experts releases technology of video compressing encoding of new generation (HEVC), expect that video compression efficiency doubles on H.264 basis again.But, due to the characteristic of human visual system (HVS) self, there is perception redundancy and be not still removed.Along with human-eye visual characteristic is studied gradually deeply, what have video worker to propose to remove human eye redundancy just can distinguish distortion model (JustNoticeableDistortion, JND).Namely according to the size of the JND threshold value tolerance perception redundancy obtained, when changing value lower than this threshold value just not by Human Perception.
Research at present for JND is mainly divided into two large class: pixel domain JND and frequency domain JND model.Wherein, the JND model proposed in document [1] is classical pixel domain model, have studied that characteristic is covered in brightness, texture covers characteristic and Temporal concealment characteristic respectively.The frequency domain JND model proposed in document [2], have studied outside first three kind characteristic, is investigated the sensitiveness of human eye to different frequency section, makes frequency domain JND model more meet the visual characteristic of human eye like this.
For the JND model proposed in document [2], it is DCT domain JND model more complete at present.It covers characteristic except the brightness comprising pixel and texture covers characteristic, also add spatial sensitivity function effect.Spatial sensitivity function reflects the bandpass characteristics of human eye, reaches removal Human Perception frequency redundancy object by removing the imperceptible frequency content of human eye.In Temporal concealment effect, contain level and smooth eyeball and move effect, not only contain the size of motion amplitude, further comprises the directional information of motion.After having researcher it to be combined with multi-view point video to act on residual error dct transform (discrete cosine transform), greatly improve compression efficiency.But, do not use it for other cataloged procedure as quantizing process, therefore its removal visual redundancy is thorough not.
The JND model set up in document [3], utilizes JND model to instruct quantizing process although propose.But its JND model set up is pixel domain, lacks the process removing human eye frequency redundancy, caused instructing quantizing process accurate not.Secondly, ensure that subjective quality for JND model, only need to carry out adjustment quantized value to the insensitive place of human eye, and other area quantization value remains unchanged.Simultaneously at adjustment quantization parameter, corresponding adjustment LaGrange parameter finally.
Patent application of the present invention proposes DCT domain JND model to be applied to quantizing process in multiple view video coding first, when ensureing that subjective quality is constant, improves video compression efficiency further.
Document [1]: X.Yang, W.Lin, andZ.Lu, " Motion-compensatedresiduepreprocessinginvideocodingbased onjust-noticeable-distortionprofile; " IEEETrans.CircuitsSyst.VideoTechnol., vol.15, no.6, pp.742 – 752,2005.
Document [2]: ZhenyuWeiandKingN.Ngan., " Spatio-TemporalJustNoticeableDistortionProfileforGreySca leImage/VideoinDCTDomain. " IEEEtransactionsoncircuitsandsystemsforvideotechnology.V OL.19, NO.3, March2009.
Document [3]: Z.ChenandC.Guillemot, " PerceptuallyfriendlyH.26/AVCvideocodingbasedonfoveatedju stnoticeabledistortionmodel; " IEEETrans.CircuitsSyst.VideoTechnol., vol.20, no.6, pp.806 – 819, Jun.2010.
Summary of the invention
The object of the invention is the defect existed for prior art, a kind of method utilizing vision perception characteristic to instruct multiple view video coding quantizing process is provided, the method is when ensureing that Subjective video quality is constant, frequency domain JND model is used to instruct multiple views quantizing process, quantization step is improved to the insensitive region of human eye, improves video compression efficiency.While adjustment step-length, the LaGrange parameter of dynamic conditioning rate-distortion optimization function, makes code efficiency improve further.
For achieving the above object, the present invention adopts following technical scheme:
Utilize vision perception characteristic to instruct a method for multiple view video coding quantizing process, it is characterized in that operating procedure is as follows:
(1) read the brightness value size of each frame of input video sequence, that sets up frequency domain just can distinguish distortion threshold model,
(2) prediction of each frame of input video sequence in viewpoint and between viewpoint,
(3) discrete cosine transform (dct transform) is carried out to residual error data,
(4) quantization step of each macro block in dynamic adjustments present frame,
(5) LaGrange parameter in dynamic adjustments rate-distortion optimization process,
(6) carry out entropy code to the data quantized, generated code flows through Internet Transmission.
The method utilizing vision perception characteristic to instruct multiple view video coding quantizing process of the present invention compared with the prior art comparatively, has following apparent outstanding substantive distinguishing features and remarkable technological progress:
1), this multi-view point video encoding method while guarantee reconstruction video mass conservation, make cataloged procedure just can reduce encoder bit rate in this subprogram by quantifying, in test, maximal rate can drop to 12.35%;
2), this multi-view point video encoding method is while guarantee reconstruction video mass conservation, adopt mean subjective mark difference, when subjective scores difference close to 0 time, illustrate that the subjective quality of two kinds of methods is more close, the mean subjective mark difference of this method is 0.03, therefore says that the subjective quality of subjective quality of the present invention and multi-view point video encoding and decoding JMVC code is suitable;
3), this multi-view point video encoding method does not increase cataloged procedure complicated especially, with less complexity raising Video coding compression efficiency.
Accompanying drawing explanation
Fig. 1 is that the vision perception characteristic that utilizes in the present invention instructs the theory diagram of the method for multiple view video coding quantizing process.
Fig. 2 is the block diagram just can distinguishing distortion model of frequency domain.
Fig. 3 is the block diagram of the intra/inter-prediction of viewpoint.
Fig. 4 is dct transform block diagram.
Fig. 5 is the block diagram of dynamic adjustments quantization step.
Fig. 6 is the block diagram of the LaGrange parameter in dynamic adjustments rate distortion costs function.
Fig. 7 is the block diagram that entropy code exports.
Fig. 8 a is the reconstruction image that video sequence ballroom the 0th viewpoint the 15th two field picture uses JMVC original coding method.
Fig. 8 b is the reconstruction image that video sequence ballroom the 0th viewpoint the 15th two field picture uses the inventive method.
Fig. 9 is that video sequence ballroom uses JMVC original coding method and the inventive method under different Q P and different points of view situation, the comparing result of code check, PSNR value, reconstruction video subjective quality assessment mark difference (DM0S).
Figure 10 a is the reconstruction image that video sequence race1 the 1st viewpoint the 35th two field picture uses JMVC original coding method.
Figure 10 b is the reconstruction image that video sequence race1 the 1st viewpoint the 35th two field picture uses the inventive method.
Figure 11 is that video sequence race1 uses JMVC original coding method and the inventive method under different Q P and different points of view situation, the comparing result of code check, PSNR value, reconstruction video subjective quality assessment mark difference (DM0S).
Figure 12 a is the reconstruction image that video sequence Crowd the 2nd viewpoint the 45th two field picture uses JMVC original coding method.
Figure 12 b is the reconstruction image that video sequence Crowd the 2nd viewpoint the 45th two field picture uses the inventive method.
Figure 13 is that video sequence Crowd uses JMVC original coding method and the inventive method under different Q P and different points of view situation, the comparing result of code check, PSNR value, reconstruction video mean subjective scoring difference (DM0S).
Embodiment
Below in conjunction with accompanying drawing, the preferred embodiments of the present invention are described in further detail:
Embodiment one:
The present embodiment utilizes vision perception characteristic to instruct the method for multiple view video coding quantizing process, see Fig. 1, comprises the following steps:
(1) read the brightness value size of each frame of input video sequence, that sets up frequency domain just can distinguish distortion threshold model,
(2) prediction of each frame of input video sequence in viewpoint and between viewpoint,
(3) discrete cosine transform is carried out to residual error data,
(4) quantization step of each macro block in dynamic adjustments present frame,
(5) LaGrange parameter in dynamic adjustments rate-distortion optimization process,
(6) carry out entropy code to the data quantized, generated code flows through Internet Transmission.
Embodiment two: the present embodiment is substantially identical with embodiment one, and special feature is as follows:
Set up frequency domain JND model in above-mentioned steps (1) and comprise four models, see Fig. 2:
(1-1) spatial contrast sensitivity function model is the bandpass characteristics curve according to human eye, for particular space frequency
its basic JND threshold value can be expressed as:
Spatial frequency
computing formula be:
Wherein,
with
represent the coordinate position of discrete cosine transform block,
for the dimension of discrete cosine transform block,
with
represent the visual angle of horizontal and vertical, it is generally acknowledged that horizontal view angle equals vertical angle of view, it is expressed as:
Because human eye vision susceptibility has directivity, more responsive to horizontal and vertical direction ratio, relatively less to the susceptibility in other directions.Add that the modulation factor in direction can obtain thus:
the angle of the frequency representated by DCT coefficient vector,
for DCT coefficient normalization factor expression formula is:
Finally add controling parameters
the modulation factor forming final spatial sensitivity function is:
In multi-vision-point encoding process, owing to there is the dct transform of 8 × 8 and 4 × 4 sizes, therefore parameter is distinguished to some extent.In an experiment, for the DCT coded format of 8 × 8 pieces of sizes,
be 0.6,
be 1.33,
be 0.11,
be 0.18; For the DCT coded format of 4 × 4 pieces of sizes,
be 0.6,
be 0.8,
be 0.035,
be 0.008.
(1-2) brightness shielding effect model is experimentally, and human eye visual perception susceptibility is in intermediate grey values region than more responsive in more black and brighter background area, and finally simulate brightness shielding effect curve, its expression formula is:
Wherein
it is the average brightness value of present encoding block.
(1-3) texture shielding effect model is the difference according to image texture, image can be divided into three regions: frontier district, smooth area and texture area.Human eye reduces its susceptibility successively.Usually canny operator is utilized to separate the regional of image.
The edge pixel density utilizing canny operator to obtain is as follows:
Wherein,
be the edge pixel sum of block, obtained by Canny edge detector.
Utilize edge pixel density
image block is divided into flat region, texture area and marginal zone, the foundation formula of image block classification is as follows:
For texture region, eyes are insensitive to low frequency part distortion, but HFS suitably retains.Therefore obtain contrasting the estimation factor covered and be:
Wherein (
) be DCT coefficient label.
Due to the eclipsing effects of spatial contrast sensitivity function effect and luminance effect, obtaining the final shielding effect factor is:
Wherein,
represent the of input video sequence
frame,
for DCT coefficient,
for the threshold value of spatial contrast degree sensitivity function,
for brightness shielding effect characteristic modulation factor.
(1-4) to be the modulation factor experimentally recording Temporal concealment effect be time contrast sensitivity function model:
Wherein,
represent temporal frequency,
representation space frequency.Temporal frequency
its general computing formula is as follows:
be respectively the horizontal and vertical component of spatial frequency,
for the speed of object of which movement on retina.
calculating formula be:
Wherein,
with
represent pixel level and vertical visual angle,
for dct transform dimension,
with
represent the coordinate position of discrete cosine transform block.
The speed of retina epigraph
computational methods are as follows:
Wherein,
be that smooth pursuit eyeball moves effect gain, in experiment, get 0.98.
represent the speed of object at the plane of delineation,
represent the minimum eyeball translational speed because drift motion causes, its empirical value is 0.15.deg/s.
be the maximal rate of the eyeball corresponding with eyes jumping, usually get 80deg/s,
it is the frame per second of video sequence.
the motion vector of each piece,
it is the visual angle of pixel.
(1-5) what namely the weighted product of four kinds of factors formed current encoded frame just can distinguish distortion threshold, and its expression formula is:
Wherein,
for the threshold value of spatial contrast degree sensitivity function,
for brightness shielding effect modulation factor,
for shielding effect modulation factor,
for Temporal concealment modulation factor.
Above-mentioned steps (2) to carry out between viewpoint/interior prediction to input video sequence, and see Fig. 3, its concrete steps are as follows:
(2-1) interframe in viewpoint/interior prediction is the time redundancy being removed present frame by the inter prediction in viewpoint, is removed the spatial redundancy of present frame by the infra-frame prediction in viewpoint.That prediction mode that selection rate aberration optimizing function is minimum in infra-frame prediction and inter prediction.Wherein rate-distortion optimization function expression is:
Wherein
for distorted signal,
for the bit number of encoding under different coding pattern,
it is the LaGrange parameter after adjustment.
(2-2) prediction carried out between viewpoint is because this method is the multiple viewpoint of coding, carries out prediction present frame, can remove the redundant information between viewpoint by the corresponding frame between viewpoint.
(2-3) to compare between viewpoint and Coding cost in viewpoint, select the best prediction mode of prediction mode again and between viewpoint to compare in viewpoint in prediction, the minimum prediction mode of selection rate aberration optimizing cost function is optimum prediction mode.To take into full account between viewpoint and redundancy properties in viewpoint, select suitable prediction mode to improve video compression efficiency further.
Above-mentioned steps (3) carries out discrete cosine transform to residual error data, and see Fig. 4, its concrete steps are as follows:
(3-1) judgement of coded block size, in multi-vision-point encoding method, coded block size has
seven kinds of situations, first four kinds are summed up as
transform block, latter three kinds are
transform block.
(3-2) corresponding dct transform, for
transform block adopts
dct transform, for
transform block adopts
dct transform.
The quantization step of each macro block in above-mentioned steps (4) dynamic adjustments present frame, see Fig. 5, its concrete steps are as follows:
(4-1) JND model by having set up, obtain the average JND value of present frame, average JND threshold value is:
Wherein,
with
the height of difference presentation video frame and width,
what represent present frame just can distinguish distortion threshold,
represent the coordinate of pixel.
(4-2) the JND average of current macro, the average JND threshold value of M macro block is expressed as:
(4-3) quantization step of dynamic adjustments current macro, just can distinguish that distortion threshold reflects the difference of human eye to the susceptibility of piece image various piece, therefore can according to just distinguishing that the difference of distortion threshold carrys out the quantization step of each macro block of dynamic adjustments.For the insensitive place of human eye, suitable for quantization step is tuned up, otherwise quantized value is constant.The quantization parameter proposed is adjusted to:
Wherein,
the original step-length of coding framework,
for regulatory factor, its expression formula is provided by following formula:
Wherein,
.
LaGrange parameter in above-mentioned steps (5) dynamic adjustments rate-distortion optimization process, see Fig. 6, its concrete operation step is as follows:
(5-1) calculate and compare the JND average of present frame and the JND average of current coding macro block, for next step provides foundation to the weighting of LaGrange parameter.
(5-2) adjust Suzanne Lenglen day parameter, before have adjusted quantization parameter, the distortion value in Lagrangian rate-distortion optimization and code check change, and now use original LaGrange parameter again
value, just can not ensure it is optimal solution.Corresponding weighting LaGrange parameter simultaneously, can make cost function again reach optimum, after adjustment
for:
Wherein,
represent the quantization parameter generated in multi-vision-point encoding method,
represent the
quantization parameter value after individual macro block adjustment.
(5-3) be updated in rate-distortion optimization cost function by the LaGrange parameter after adjustment, its expression formula is as follows:
Wherein
for distorted signal,
for the bit number of encoding under different coding pattern,
it is the LaGrange parameter after adjustment.Make like this while quantization parameter changes, corresponding change LaGrange parameter, makes rate-distortion optimization function still obtain optimal solution.
Above-mentioned steps (6) carries out entropy code to the data quantized, and generated code flows through Internet Transmission, and see Fig. 7, its concrete steps are as follows:
(6-1) entropy code is carried out to the data quantized, make the data quantized the most effectively can be represented by binary code stream like this, eliminate the statistical redundancy of quantized data.
(6-2) code stream formed by entropy code, by Internet Transmission, realizes the transmission of video.Through vision perception characteristic process coding method due to its occupied bandwidth little, better can adapt to Internet Transmission.
Carry out a large amount of emulation experiment below to assess the proposed performance utilizing the multi-view point video encoding method of visual characteristic.Be configured to IntelPentium4CPU3.00GHz, front 48 frames of encoding and decoding multi-view video sequences ballroom, race1, crowd on the PC of 512MInternalMemory, Intel8254GExpressChipsetFamily, WindowsXPOperationSystem, wherein, BASICQP is set to 20, and 24,28,32, experiment porch selects multi-view point video encoding and decoding reference software JMVC, and encoding and decoding predict selects HHI-IBBBP, and interview prediction mode adopts bi-directional predicted mode.
The experimental result of video sequence ballroom is as shown in Fig. 8 a ~ 8b, Fig. 9.Fig. 8 a be video sequence ballroom when quantization parameter QP=24, the 0th viewpoint the 15th two field picture uses the reconstruction image of JMVC original coding method, the PSNR=40.31dB of reconstruction video image.Fig. 8 b be video sequence ballroom when quantization parameter QP=24, the 0th viewpoint the 15th two field picture uses the reconstruction video image of the inventive method, the PSNR=40.10dB of reconstruction video image.Fig. 9 is that video sequence ballroom uses JMVC original coding and the present invention's two kinds of methods, when different Q P and different points of view, code check, PSNR value, code check save the statistics of percentage, reconstruction video subjective quality assessment mark difference (DM0S), average bit rate saving percentage.Can find out, video sequence ballroom is under different Q P, use the encoder bit rate of the inventive method than using the encoder bit rate of JMVC original coding method to save 7.47% ~ 9.16%, the Subjective video quality evaluation score difference of JMVC original coding method and the inventive method is 0.03 ~ 0.07, can think that subjective quality remains unchanged.
The experimental result of video sequence race1 is as shown in Figure 10 a ~ 10b, Figure 11.Figure 10 a be video sequence race1 when quantization parameter QP=24, the 1st viewpoint the 25th two field picture uses the reconstruction video image of JMVC original coding method, the PSNR=41.15dB of reconstruction video image.Figure 10 b be video sequence race1 when quantization parameter QP=24, the 1st viewpoint the 36th two field picture uses the reconstruction video image of JMVC original coding method, the PSNR=40.51dB of reconstruction video image.Figure 11 is that video sequence race1 uses JMVC original coding and the present invention's two kinds of methods, when different Q P and different points of view, code check, PSNR value, code check save the statistics of percentage, reconstruction video subjective quality assessment mark difference (DM0S), average bit rate saving percentage.Can find out, video sequence race1 is under different Q P, use the encoder bit rate of the inventive method than using the encoder bit rate of JMVC original coding method to save 10.77% ~ 12.35%, the Subjective video quality evaluation score difference of JMVC original coding method and the inventive method is 0.06 ~ 0.09, can think that subjective quality remains unchanged.
The experimental result of video sequence crowd is as shown in Figure 12 a ~ 12b, Figure 13.Figure 12 a be video sequence crowd when quantization parameter QP=35, the 2nd viewpoint the 45th two field picture uses the reconstruction video image of JMVC original coding method, the PSNR=33.77dB of reconstruction video image.Figure 12 b be video sequence crowd when quantization parameter QP=35, the 2nd viewpoint the 45th two field picture uses the reconstruction video image of JMVC original coding method, the PSNR=33.12dB of reconstruction video image.Figure 13 is that video sequence crowd uses JMVC original coding and the present invention's two kinds of methods, when different Q P and different points of view, code check, PSNR value, code check save the statistics of percentage, reconstruction video subjective quality assessment mark difference (DM0S), average bit rate saving percentage.Can find out, video sequence crowd is under different Q P, use the encoder bit rate of the inventive method than using the encoder bit rate of JMVC original coding method to save 8.95% ~ 9.83%, the Subjective video quality evaluation score difference of JMVC original coding method and the inventive method is 0.03 ~ 0.08, can think that subjective quality remains unchanged.
Can find out in conjunction with above each chart, the present invention is by setting up the JND model of DCT domain, and applied to multiple view video coding framework quantizing process and rate-distortion optimization process, when ensureing that subjective quality is constant, significantly reduce multiple view video coding code check, improve the compression efficiency of multiple view video coding.
Claims (6)
1. utilize vision perception characteristic to instruct a method for multiple view video coding quantizing process, it is characterized in that operating procedure is as follows:
(1) read the brightness value size of each frame of input video sequence, that sets up frequency domain just can distinguish distortion threshold model,
(2) prediction of each frame of input video sequence in viewpoint and between viewpoint,
(3) discrete cosine transform is carried out to residual error data,
(4) quantization step of each macro block in dynamic adjustments present frame,
(5) LaGrange parameter in dynamic adjustments rate-distortion optimization process, concrete steps are:
1. the average just can distinguishing distortion threshold just can distinguishing distortion threshold average and current macro of more every frame;
2. adjust LaGrange parameter, its expression formula of the LaGrange parameter after adjustment is:
Wherein
for regulatory factor,
for the quantization step after adjustment,
the original quantization step of presentation code framework,
represent the average just can distinguishing distortion threshold of current macro,
represent the average just can distinguishing distortion threshold of present frame;
3. the optimization of Coding cost function, dynamic adjustments LaGrange parameter, makes rate-distortion optimization function when quantization step changes, regains optimal solution; Its expression formula is:
Wherein
for distorted signal,
for the bit number of encoding under different coding pattern,
it is the LaGrange parameter after adjustment;
(6) carry out entropy code to the data quantized, generated code flows through Internet Transmission.
2. the method utilizing vision perception characteristic to instruct multiple view video coding quantizing process according to claim 1, it is characterized in that described step (1) reads the brightness value size of each frame of input video sequence, that sets up frequency domain just can distinguish that the operating procedure of distortion threshold model is as follows:
1. the spatial sensitivity factor of 4x4 and 8x8DCT conversion is obtained respectively according to the dimension of dct transform
, its formula is:
Wherein s is controling parameters,
the angle of the frequency representated by DCT coefficient vector,
for DCT coefficient normalization factor expression formula is:
,
for spatial frequency, parameter r, a, b and c are different according to varying in size of dct transform: for the DCT coded format of 8 × 8 pieces of sizes,
be 0.6,
be 1.33,
be 0.11,
be 0.18; For the DCT coded format of 4 × 4 pieces of sizes,
be 0.6,
be 0.8,
be 0.035,
be 0.008;
2. human eye is experimentally recorded under Different background illumination condition, brightness shielding effect
curve is expressed as follows:
Wherein,
for the average pixel value of present encoding block;
3. utilize edge detector to detect the texture features of present encoding block, obtain texture and cover the factor
, its expression formula is as follows:
Wherein,
represent the transverse and longitudinal coordinate coefficient of transform block,
represent that the estimation factor is covered in contrast,
for the spatial sensitivity factor,
for the dct transform coefficient of the n-th encoding block of present frame;
4. according to the speed of object of which movement in the every frame of video sequence, experiment records Temporal concealment effector
expression formula is:
Wherein,
for spatial frequency,
for temporal frequency;
5. described step 1. ~ weighted product of four kinds of factors of 4. trying to achieve namely form current encoded frame just can distinguish distortion threshold.
3. the method utilizing vision perception characteristic to instruct multiple view video coding quantizing process according to claim 1, is characterized in that the operating procedure of the prediction of each frame of described step (2) input video sequence in viewpoint and between viewpoint is as follows:
1. carry out the interframe in viewpoint and infra-frame prediction, predicted value and the current frame that will encode are compared, chooses a kind of coded system that Coding cost is less;
2. carry out the prediction between viewpoint, the current encoded frame of current view point is predicted according to the corresponding frame of reference view, is compared by the corresponding frame of predicted value and reference view, tries to achieve the Coding cost of interview prediction;
3. to compare between viewpoint and Coding cost in viewpoint, select that predictive mode compared with lower Item cost.
4. the method utilizing vision perception characteristic to instruct multiple view video coding quantizing process according to claim 1, is characterized in that the operating procedure that described step (3) carries out discrete cosine transform to residual error data is as follows:
1. the judgement of coded block size, when arbitrary length of side of encoding block is less than 8, then classifies as 4x4 transform block, otherwise, be then 8x8 transform block;
2. when being 4x4 transform block, selecting 4x4DCT conversion, when being 8x8 transform block, selecting 8x8DCT conversion.
5. the method utilizing vision perception characteristic to instruct multiple view video coding quantizing process according to claim 1, is characterized in that the operating procedure of the quantization step of each macro block in described step (4) dynamic adjustments present frame is as follows:
1. the mean value just can distinguishing distortion threshold of present frame is calculated;
What 2. calculate current coding macro block just can distinguish distortion threshold mean value;
3. the average just can distinguishing distortion threshold just can distinguishing distortion threshold average and current macro of more every frame, the quantization step of dynamic adjustments current macro, its expression formula of the quantization step after adjustment is as follows:
Wherein,
the original quantization step of presentation code framework,
represent the average just can distinguishing distortion threshold of current macro,
represent the average just can distinguishing distortion threshold of present frame,
for regulatory factor.
6. the method utilizing vision perception characteristic to instruct multiple view video coding quantizing process according to claim 1, it is characterized in that described step (6) carries out entropy code to the data quantized, the operating procedure that generated code flows through Internet Transmission is as follows:
1. the data after quantizing carry out entropy code, make the data after quantizing form binary code stream;
2. encoding code stream passes through Internet Transmission.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201210402003.9A CN103124347B (en) | 2012-10-22 | 2012-10-22 | Vision perception characteristic is utilized to instruct the method for multiple view video coding quantizing process |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201210402003.9A CN103124347B (en) | 2012-10-22 | 2012-10-22 | Vision perception characteristic is utilized to instruct the method for multiple view video coding quantizing process |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN103124347A CN103124347A (en) | 2013-05-29 |
| CN103124347B true CN103124347B (en) | 2016-04-27 |
Family
ID=48455183
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201210402003.9A Expired - Fee Related CN103124347B (en) | 2012-10-22 | 2012-10-22 | Vision perception characteristic is utilized to instruct the method for multiple view video coding quantizing process |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN103124347B (en) |
Families Citing this family (21)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104488266B (en) * | 2013-06-27 | 2018-07-06 | 北京大学深圳研究生院 | AVS video compressing and encoding methods and encoder |
| CN103475875B (en) * | 2013-06-27 | 2017-02-08 | 上海大学 | Image adaptive measuring method based on compressed sensing |
| CN103716623B (en) * | 2013-12-17 | 2017-02-15 | 北京大学深圳研究生院 | Video compression encoding-and-decoding method and encoder-decoder on the basis of weighting quantification |
| EP2887664A1 (en) * | 2013-12-19 | 2015-06-24 | Thomson Licensing | Method and device for encoding a high-dynamic range image |
| CN104219526B (en) * | 2014-09-01 | 2017-05-24 | 国家广播电影电视总局广播科学研究院 | HEVC rate distortion optimization algorithm based on just-noticeable perception quality judging criterion |
| CN104349167B (en) * | 2014-11-17 | 2018-01-19 | 电子科技大学 | A kind of method of adjustment of Video coding rate-distortion optimization |
| CN104469386B (en) * | 2014-12-15 | 2017-07-04 | 西安电子科技大学 | A kind of perception method for encoding stereo video of the proper appreciable error model based on DOF |
| CN105245890B (en) * | 2015-10-16 | 2018-01-19 | 北京工业大学 | A kind of efficient video coding method of view-based access control model attention rate priority |
| CN105704497B (en) * | 2016-01-30 | 2018-08-17 | 上海大学 | Coding unit size fast selection algorithm towards 3D-HEVC |
| CN105812805A (en) * | 2016-01-31 | 2016-07-27 | 西安电子科技大学 | Coding method and apparatus for video images |
| CN106454386B (en) * | 2016-10-26 | 2019-07-05 | 广东电网有限责任公司电力科学研究院 | A kind of method and apparatus of the Video coding based on JND |
| CN108574841B (en) * | 2017-03-07 | 2020-10-30 | 北京金山云网络技术有限公司 | Coding method and device based on self-adaptive quantization parameter |
| CN107094251B (en) * | 2017-03-31 | 2021-07-23 | 浙江大学 | Video and image coding and decoding method and device based on spatial position adaptive quality adjustment |
| CN107197266B (en) * | 2017-06-26 | 2020-05-15 | 杭州当虹科技股份有限公司 | HDR video coding method |
| US10614609B2 (en) * | 2017-07-19 | 2020-04-07 | Mediatek Inc. | Method and apparatus for reduction of artifacts at discontinuous boundaries in coded virtual-reality images |
| CN108111852B (en) * | 2018-01-12 | 2020-05-29 | 东华大学 | Double-measurement-parameter rate-distortion control method for quantization block compressed sensing coding |
| CN110113606A (en) * | 2019-03-12 | 2019-08-09 | 佛山市顺德区中山大学研究院 | A kind of method, apparatus and equipment of removal human eye perception redundant video coding |
| WO2021078653A1 (en) * | 2019-10-21 | 2021-04-29 | Fraunhofer-Gesellschaft zur Förderung der angewandten Forschung e.V. | Weighted psnr quality metric for video coded data |
| CN113489983A (en) * | 2021-06-11 | 2021-10-08 | 浙江智慧视频安防创新中心有限公司 | Method and device for determining block coding parameters based on correlation comparison |
| CN114554219A (en) * | 2022-02-21 | 2022-05-27 | 翱捷科技股份有限公司 | Rate distortion optimization method and device based on motion detection |
| CN115967806B (en) * | 2023-03-13 | 2023-07-04 | 阿里巴巴（中国）有限公司 | Data frame coding control method, system and electronic equipment |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1968419A (en) * | 2005-11-16 | 2007-05-23 | 三星电子株式会社 | Image encoding method and apparatus and image decoding method and apparatus using characteristics of the human visual system |
| CN101710995A (en) * | 2009-12-10 | 2010-05-19 | 武汉大学 | Video coding system based on vision characteristic |
| CN101854555A (en) * | 2010-06-18 | 2010-10-06 | 上海交通大学 | Video coding system based on prediction residual self-adaptation regulation |
| CN102420988A (en) * | 2011-12-02 | 2012-04-18 | 上海大学 | Multi-view video coding system utilizing visual characteristics |
| CN102724525A (en) * | 2012-06-01 | 2012-10-10 | 宁波大学 | Depth video coding method on basis of foveal JND (just noticeable distortion) model |
-
2012
- 2012-10-22 CN CN201210402003.9A patent/CN103124347B/en not_active Expired - Fee Related
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1968419A (en) * | 2005-11-16 | 2007-05-23 | 三星电子株式会社 | Image encoding method and apparatus and image decoding method and apparatus using characteristics of the human visual system |
| CN101710995A (en) * | 2009-12-10 | 2010-05-19 | 武汉大学 | Video coding system based on vision characteristic |
| CN101854555A (en) * | 2010-06-18 | 2010-10-06 | 上海交通大学 | Video coding system based on prediction residual self-adaptation regulation |
| CN102420988A (en) * | 2011-12-02 | 2012-04-18 | 上海大学 | Multi-view video coding system utilizing visual characteristics |
| CN102724525A (en) * | 2012-06-01 | 2012-10-10 | 宁波大学 | Depth video coding method on basis of foveal JND (just noticeable distortion) model |
Also Published As
| Publication number | Publication date |
|---|---|
| CN103124347A (en) | 2013-05-29 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN103124347B (en) | Vision perception characteristic is utilized to instruct the method for multiple view video coding quantizing process | |
| CN102420988B (en) | Multi-view video coding system utilizing visual characteristics | |
| CN102186070B (en) | Method for realizing rapid video coding by adopting hierarchical structure anticipation | |
| CN100566411C (en) | Eliminate method, medium and the filter of blocking effect | |
| CN102611892B (en) | Method and device for realizing adaptive quantization in coding process | |
| CN106534862B (en) | Video coding method | |
| CN103369349B (en) | A kind of digital video-frequency quality control method and device thereof | |
| CN103079063B (en) | A kind of method for video coding of vision attention region under low bit rate | |
| CN106937116B (en) | Low-complexity video coding method based on random training set adaptive learning | |
| CN104219525B (en) | Perception method for video coding based on conspicuousness and minimum discernable distortion | |
| CN107211145A (en) | The almost video recompression of virtually lossless | |
| JP2008503918A5 (en) | ||
| CN104539962A (en) | Layered video coding method fused with visual perception features | |
| WO2010078759A1 (en) | Method for image temporal and spatial resolution processing based on code rate control | |
| CN101710993A (en) | Block-based self-adaptive super-resolution video processing method and system | |
| CN101184221A (en) | Vision attention based video encoding method | |
| CN101325711A (en) | Method for controlling self-adaption code rate based on space-time shielding effect | |
| CN107241607B (en) | Visual perception coding method based on multi-domain JND model | |
| CN108924554A (en) | A kind of panorama video code Rate-distortion optimization method of spherical shape weighting structures similarity | |
| CN101272489B (en) | Encoding and decoding device and method for video image quality enhancement | |
| CN103051901A (en) | Video data coding device and video data encoding method | |
| CN106303521B (en) | A kind of HEVC Rate-distortion optimization method based on sensitivity of awareness | |
| CN109982082B (en) | HEVC multi-distortion criterion rate-distortion optimization method based on local texture characteristics | |
| CN111083477A (en) | HEVC (high efficiency video coding) optimization algorithm based on visual saliency | |
| CN105657433A (en) | Image complexity based signal source real-time coding method and system |
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: 20160427 Termination date: 20211022 |
|
| CF01 | Termination of patent right due to non-payment of annual fee |