CN103546749A - Method for optimizing HEVC (high efficiency video coding) residual coding by using residual coefficient distribution features and bayes theorem - Google Patents
Method for optimizing HEVC (high efficiency video coding) residual coding by using residual coefficient distribution features and bayes theorem Download PDFInfo
- Publication number
- CN103546749A CN103546749A CN201310480055.2A CN201310480055A CN103546749A CN 103546749 A CN103546749 A CN 103546749A CN 201310480055 A CN201310480055 A CN 201310480055A CN 103546749 A CN103546749 A CN 103546749A
- Authority
- CN
- China
- Prior art keywords
- residual
- coding
- encoding
- block
- 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.)
- Granted
Links
Images
Abstract
The invention relates to a method for optimizing residual coding by using residual coefficient distribution features and bayes theorem. The method comprises the following steps of (1) reading each frame of images in a video sequence according to a sequence set in a control file; (2) performing intra-frame prediction and inter-frame prediction on the brightness and the chromatic value of each frame of images so as to obtain a residual coefficient; (3) performing texture judgment or bayes model judgment on a residual block TU (the maximum is 32*32) according to the residual coefficient so as to judge whether block encoding is required to be stopped ahead of time or not; (4) performing DCT (discrete cosine transform) and quantification on the whole residual block TU so as to obtain a quantification parameter; and (5) performing entropy encoding on the quantification coefficient, and finally outputting the quantification coefficient in a bit stream mode. By the method, the video encoding speed is increased under the condition that the loss on the encoding quality can be ignored, and videos can be conveniently acquired in real time.
Description
Technical field
The present invention relates to high-resolution video coding techniques field, particularly utilize residual error coefficient distribution characteristics and Bayes' theorem to optimize the method for HEVC residual coding, be applicable to high-resolution video coding and real time video collection.
Background technology
In recent years due to the fast development of multimedia technology, HDTV (High-Definition Television) (HDTV), 3D three-dimensional video-frequency, the new technologies such as video communication are well known gradually, and people have been brought into brand-new video epoch.All these technology, when offering the multimodal various high-quality video enjoyment of people, are also had higher requirement to video coding technique.Wherein, most important feature is exactly will process more googol according to amount than these new technologies of SD two-dimensional video before, existing video encoding standard, as H.264 unable to do what one wishes, applied and is given birth to so have the more video encoding standard HEVC of high compression efficiency.For improving code efficiency HEVC, adopted many new technology, as larger encoding block, circular recursion coding structure, more intra prediction mode etc., these improve when improving code efficiency, also make the whole algorithm very complex that becomes, and are unfavorable for that very much it drops into practical application.For this reason, how effectively to reduce HEVC encoder complexity and become a study hotspot now.
Infra-frame prediction, inter prediction have taken most of the time in whole video coding process, become main direction of studying, and method has been tending towards ripe.And the residual coding that comprises integer transform, quantification and inverse transformation and inverse quantization has also taken many scramble times, some scholars are studied it.Wherein, document [1] is being analyzed HEVC residual coding piece TU(Transform Unit) tree structure and circular recursion basis of coding on proposed a kind of easy TU and ended in advance scheme.Nonzero coefficient number in quantization parameter by statistics residual block through integer transform and after quantizing, determines whether ending block encoding.The scramble time that the method is saved is a lot, but coding quality loss is also larger.
The HEVC residual coding optimized algorithm that document [2] proposes is fairly perfect.First author has analyzed optimum code mode decision standard: rate distortion costs function, and then find out monoblock TU and four sub-block TU thereof the correlation on coding bit rate, and analyzed the mutual derivation relation of complete zero block type between monoblock TU and its sub-block.Rate distortion costs by contrast TU tetra-sub-blocks and with the size of predetermined threshold value, realize and skip bulk residual coding and end fritter residual coding.
Document [3] is for the residual coding difference of HEVC infra-frame prediction and inter prediction, proposed respectively about the fast coding algorithm of infra-frame prediction with about the quick residual coding algorithm of inter prediction.On the basis of the former original scheme in identifying code HM2.0, further reduce the intra prediction mode number of participation rate distortion computation in infra-frame prediction, but every kind of pattern is all carried out to the residual error tree search of limit, thereby on the basis of few scramble time, also guaranteed coding quality falling.First the latter finds out maximum and the minimum value in the parameter coefficient of four sub-TU and makes them poor, thereby then this difference is compared with setting threshold and determined whether and will end in advance piecemeal residual coding.
Document [1]: Kiho Choi and Euee S.Jang, " Early TU decision method for fast video encoding in high efficiency video coding, " ELECTRONICS LETTERS, Vol. 48, No. 12,7th June, 2012.
Document [2]: Su-Wei Teng, Hsueh-Ming Hang and Yi-Fu Chen, " Fast Mode Decision Algorithm for Residual Quadtree Coding in HEVC; " IEEE Visual Communications and Image Processing.VCIP.2011.6116062, pp. 1-4,2011.
Document [3]: Yih Han Tan, Chuohao Yeo, Hui Li Tan and Zhengguo Li, " On Residual Quad-tree Coding In HEVC; " IEEE Multimedia Signal Processing, 13th International Workshop. MMSP.2011.6093805, pp. 1 – 4,2011.
summary of the invention
The object of the invention is for the distinctive residual coding structure of HEVC and technological deficiency, a kind of method of utilizing residual error coefficient branch feature and Bayes' theorem to optimize HEVC residual coding is provided, in the situation that guaranteeing that subjective quality is constant, can effectively improve Video coding speed.For achieving the above object, design of the present invention is: first, guaranteeing that video encoding quality changes under little prerequisite, utilize the gaussian distribution characteristic of residual error coefficient, upper and lower two sub-blocks of residual block TU and two of left and right sub-block are carried out to twice hypothesis testing, thereby judge the inner vein characteristic of current residual block TU, to reduce unnecessary sub-block residual coding.The second, utilize Bayes decision model, TU is ended to subblock coding in advance coding rate is improved.
According to foregoing invention design, the present invention adopts following technical scheme:
Utilize residual error coefficient distribution characteristics and Bayes' theorem to optimize a method for HEVC residual coding, it is characterized in that operating procedure is as follows:
(1) input video sequence: according to the order of setting in control documents, read each two field picture of video sequence,
In frame (2)/inter prediction: the brightness of every two field picture and chromatic value are carried out to infra-frame prediction and inter prediction, thereby try to achieve residual error coefficient,
(3) TU ends to judge in advance: according to residual error coefficient, residual block TU(is to the maximum to 32 * 32) carry out texture judgement or carry out Bayesian model judgement, thus determine whether to end in advance block encoding,
(4) dct transform and quantification: monoblock TU is carried out to dct transform and quantification, thereby try to achieve quantization parameter,
(5) entropy coding: quantization parameter is carried out to entropy coding, the last formal output with bit stream.
The present invention compared with the prior art, has following apparent outstanding substantive distinguishing features and significantly technological progress:
1), this HEVC video encoding optimization method when guaranteeing that video encoding quality is constant, make cataloged procedure just can improve coding rate at this subprocess of residual coding, the residual coding time that can save in experiment mostly is 60% most;
2) optimized algorithm, about residual error coefficient distribution character in this HEVC video encoding optimization scheme detects based on hypothesis, so can realize the compromise between coding quality and coding rate by revising the level of signifiance according to the actual requirements;
3) the mean square deviation MAD that, in this HEVC video encoding optimization scheme, the optimized algorithm based on Bayes decision model is chosen predicated error and residual error coefficient is that the characteristic vector factor is carried out mode decision, and these two values are just obtained in frame and in inter prediction, so this optimization method can not increase the scramble time extraly.
Accompanying drawing explanation
Fig. 1 is that utilize residual error coefficient distribution characteristics and the Bayes' theorem in the present invention optimized the theory diagram of the method for HEVC residual coding.
Fig. 2 is the block diagram of Skip, interframe and infra-frame prediction.
Fig. 3 is based on residual error coefficient distribution characteristics algorithm block diagram.
Fig. 4 is the structured flowchart of TU.
Fig. 5 is when different CU, PU and interframe and infra-frame prediction, can be used for the TU type of residual coding.
Fig. 6 is entropy coding schematic block diagram.
Fig. 7 a is that resolution is 832 * 480 the cycle tests RaceHorses RD curve chart under the optimized algorithm based on residual error coefficient distribution characteristics.
Fig. 7 b is that resolution is 1280 * 720 the cycle tests vidvo1 RD curve chart under the optimized algorithm based on residual error coefficient distribution characteristics.
Fig. 8 is the experimental result of comparing with original method in HM5.0 based on residual error coefficient distribution characteristics optimized algorithm, and major parameter comprises: the PSNR of brightness, bit rate and residual coding time.
Fig. 9 a is that resolution is 832 * 480 the cycle tests RaceHorses RD curve chart under the optimized algorithm based on Bayes decision model.
Fig. 9 b is that resolution is 1280 * 720 the cycle tests vidvo1 RD curve chart under the optimized algorithm based on Bayes decision model.
Figure 10 is the experimental result of comparing with original method in HM5.0 based on Bayes decision model optimization algorithm, and major parameter comprises: the PSNR of brightness, bit rate and residual coding time.
Embodiment
Below in conjunction with accompanying drawing, the preferred embodiments of the present invention are described in further detail:
Embodiment mono-:
Utilize the distribution characteristics of residual block coefficient and the method (referring to Fig. 1) of Bayes' theorem optimization HEVC residual coding to comprise the following steps:
(1) input video sequence: according to the order of setting in control documents, read each two field picture of video sequence,
In frame (2)/interframe: the brightness of every two field picture and chromatic value are carried out to infra-frame prediction and inter prediction, thus the residual error of trying to achieve
Coefficient,
(3) TU ends to judge in advance: according to residual error coefficient, residual block TU(is to the maximum to 32 * 32) carry out texture judgement or
Carry out Bayesian model judgement, thereby determine whether to end in advance block encoding,
(4) dct transform and quantification: monoblock TU is carried out to dct transform and quantification, thereby try to achieve quantization parameter,
(5) entropy coding: quantization parameter is carried out to entropy coding, the last formal output with bit stream.
Embodiment bis-: the present embodiment and embodiment mono-are basic identical, and special feature is as follows: (seeing Fig. 2 to Figure 10)
Above-mentioned steps (2) is to carry out in frame and inter prediction to input video sequence, and referring to Fig. 2, its concrete steps are as follows:
(2-1) infra-frame prediction be utilize around encoding block (left side and top) thus coded reference pixel has been predicted and has been eliminated video image redundancy spatially current block; Thereby inter prediction carries out to it time redundancy that Motion estimation and compensation is eliminated video sequence by reference to the front and back frame of encoding block place frame.HEVC according to the criterion of rate distortion optimum in a plurality of frames and select in optimum frame and inter-frame forecast mode interframe candidate pattern.Rate distortion costs function is:
Wherein
for rate distortion costs value,
for predicted distortion value,
for the bit number of exporting under different predictive modes,
for LaGrange parameter.
(2-2) H.264 different from video encoding standard before, for improving code efficiency, HEVC adopts larger encoding block (64 * 64), the coded system of circular recursion, intra prediction mode is increased to 35 kinds, and these improve and when improving code efficiency, also make to become more complicated with inter prediction in frame.
If (2-3) input is I frame, only it is carried out infra-frame prediction and carries out rate-distortion optimization; If P frame or B frame, first carry out Skip model prediction and inter prediction to it, and then carry out infra-frame prediction.Finally by rate-distortion optimization standard, from three kinds of patterns, select optimum predictive mode.
The method that above-mentioned steps (3) is judged the coefficient of residual block has two kinds, and concrete steps are as follows:
(3-1) hypothesis test based on Gaussian Profile
From pertinent literature research, it is zero Gaussian Profile that residual error coefficient is obeyed desired value, meets following formula:
（2）
Wherein
for average,
for variance.
For judging whether monoblock TU is applicable to being divided into four sub-TU and monoblock TU can be divided into two kinds of dividing mode, as shown in Figure 3.If two kinds of dividing mode all meet, can think that TU is applicable to being divided into four sub-TU.If the dividing mode encoding efficiency in employing Fig. 3 is better, residual error coefficient should Gaussian distributed, owing to obeying the desired value of two parts of same distribution
should there is no difference.Therefore, we can judge by hypothesis testing and the prediction effect of this pattern obtain according to the knowledge of probability theory:
In above formula
be distributed as Xue Shengshi and distribute,
the degree of freedom of this distribution,
,
the average of two parts pixel intensity in every kind of dividing mode,
,
two parts mathematical expectations separately, in above formula
as follows:
Obviously, 32 * 32 macro blocks are divided into two equal parts, the number of pixels of two parts should equate, that is:
.According to hypothesis testing, should meet, get significance level=0.05 simultaneously, looking into t distribution table can obtain
, we can obtain as drawn a conclusion:
Above formula is equivalent to:
Above formula is to calculate for 32 * 32 TU.In HEVC residual coding, TU adopts and the similar quadtree coding structure of CU, except 32 * 32, also to carry out traversal formula coding to its sub-block 16 * 16,8 * 8 and 4 * 4, then from all patterns, select optimization model, we can carry out 16 * 16 and 8 * 8 TU similarly processing with 32 * 32 for this reason, thereby obtain
（7）
Above-mentioned three formulas are respectively to calculate the Rule of judgment of 32 * 32,16 * 16 and 8 * 8.If meet formula (6), can think not significantly difference of this kind of division in 32 * 32TU, if meet formula (7), (8), can think 16 * 16, the not significantly difference of this kind of division in 8 * 8TU.Residual error coefficient is relevant with coding mode, and simple for what calculate, we only.Based on above analysis, treatment step in this paper is as follows:
1) first, adopt the monoblock TU to 32 * 32 to carry out residual coding.
2) when all there is no significantly difference according to two kinds of dividing mode shown in Fig. 3, redirect 3); Otherwise the TU degree of depth adds 1, and return to 1).
3) monoblock TU is carried out to piecemeal residual coding.
(3-2) decision algorithm based on bayesian theory
The TU of a certain size carries out block encoding and non-block encoding is two events of opposition completely, by this event definition of TU block encoding, is
, and TU does not carry out block encoding, be not defined as
.TU to measure feature
by the accuracy of helping improve classification.
that TU characteristic vector is divided into while being F
the posterior probability of class.In residual coding process, if make the decision making mistake, should block encoding in fact do not have a block encoding, and original block encoding that should piecemeal, this will cause rate distortion loss.We are should piecemeal and actually do not have the loss that piecemeal causes to be labeled as
, and original should piecemeal in fact piecemeal the loss that causes be labeled as
.Between them, there is following relation:
Wherein,
with
it is the rate distortion costs producing when TU carries out block encoding and non-block encoding.As shown from the above formula when TU makes selecting properly, should block encoding in fact also piecemeal or not piecemeal in fact really do not have piecemeal yet, will can not cause any rate distortion loss, therefore
.Based on above analysis, can obtain artificial situation
time Bayes risk cost
:
（12）
When
<
time, select
the cost causing is less, should end TU to carry out block encoding.And work as
>
time, select
the cost causing is less, should to TU, carry out block encoding according to original algorithm.In above formula:
Wherein
while representing that TU carries out Splite and None-Splite coding, the probability density function of its characteristic vector F,
it is the priori probability density function of situation Splite and None-Splite.By formula (11), (12) and (13), can be obtained that whether TU is carried out to the judgement formula of block encoding is as follows:
For the probability distribution effectively doping, herein using the mean square deviation MAD of residual block coefficient and and the predicated error of current TU piece as the key element to measure feature F.The non-parametric density estimation technique proposing according to people such as D.Chai is tried to achieve conditional probability density function
and be placed in a question blank.For reducing statistical work amount, two characteristic vector key elements can be quantized into 10 deciles, whole like this characteristic vector F will be divided into 100 scales.And in (14) formula in the decision threshold of the inequality left side
,
,
with
all relevant with the degree of depth of resolution, QP size and the TU of video.Therefore, in advance the video sequence of one group of different resolution is added up under different QP herein, thereby obtained the decision threshold under difference minute variability, different Q P and the different TU degree of depth, and these threshold values are placed in another one question blank.
In sum, the TU type decision method based on bayesian theory comprises following step:
1) decision threshold on inequality the right in (14) formula is carried out to initialization, thus obtain can be to video resolution and
The decision threshold of QP adaptive change.
2) to a TU piece from its root node (32 * 32), monoblock TU is carried out to residual coding.Obtain the MAD. of encoding error and residual error coefficient
3) by the MAD of monoblock encoding error and residual error coefficient table look-up find out TU to measure feature F probability-distribution function
with
.By (14) formula, judge, if inequality meets the demands, jump to (4).Otherwise TU is divided into four identical sub-TU, depth D epth adds 1, forwards (2) to.
4) by the order of lining by line scan, next TU is processed equally.
Above-mentioned steps (4) is carried out integer transform (DCT) and quantizes residual error data, and its concrete steps are as follows:
(4-1) when HEVC carries out integer transform and quantizes, as shown in Figure 4, the full-size of TU is 32 * 32 to the coding structure of TU, and minimum dimension is 4 * 4, is similar to the quad-tree structure of CU.
(4-2) in different CU, PU, frame and inter prediction, TU has different available types, as shown in Figure 5.Wherein, the TU that can use when red italic represents infra-frame prediction, and in table, listed all patterns all can be used when inter prediction.
Above-mentioned steps (5) is carried out entropy coding to the quantization parameter obtaining through residual coding, finally form binary bit stream with
Be convenient to transmit in network.Referring to Fig. 7, its concrete steps are as follows:
(5-1) quantization parameter is carried out to variable entropy coding (VLC) or arithmetic coding (CABAC), thereby the symbol redundancy of remove quantization coefficient realizes the further compression to video sequence.
(5-2) through entropy coded data finally with the formal output of bit stream.The adaptively changing that can realize bit rate by relevant code check control technology, this has improved the network friendliness of HEVC encoder greatly.
For verifying, in this paperly based on residual error coefficient distribution character and two kinds of quick residual block types based on bayesian theory, sentence
Determine the validity of method, below to a large amount of cycle testss have been carried out to emulation experiment.Experiment porch (PC) is configured to: Intel Core2 Duo CPU, 2.53 GHz, 1G Internal Memory, Windows XP Operation System; Standard Knowledge Verification Model is HEVC canonical reference code HM 5.0; Adopt YC to separate coded system (only brightness being optimized) herein; Emulation experiment is that front 20 frames of 416 * 240,832 * 480,1280 * 720 and 1920 * 1080 cycle tests are added up to resolution, and wherein BASIC QP is made as 22,27,32,37; Encoding and decoding predict is selected HHI-IPPP.
Method based on residual error coefficient distribution character (Proposed A) experimental result is as shown in Fig. 7 a～7b, Fig. 8.Fig. 7 a, 7b are respectively that resolution is 832 * 480 and 1280 * 720 video sequence PartyScene and the RD curve chart of vidyo1 under the coding method of HM5.0 raw residual and two kinds of methods of Proposed A.Can find out that in Fig. 7 a and Fig. 7 b, two curves almost completely overlap, the coding quality of the method based on parameter Coefficients Distribution in this explanation the present invention is almost identical with original method in HM5.0.Fig. 8 is the result that HEVC raw residual coding and two kinds of methods of Proposed A are tested the video sequence of different resolution in the situation that of different Q P, and leading indicator has DPSNR, DBR and DTime, is respectively customized for:
Wherein, PSNR is Y-PSNR, and BR is video code bit rate, and Time is the residual coding time.Can find out that the optimization method residual coding time average in the present invention has reduced approximately 42% in the situation that guaranteeing that coding quality is constant (coding bit rate on average reduces by 0.15%, and the PSNR of brightness on average reduces 0.024dB).
Quick residual coding method (Proposed B) experimental result based on bayesian theory is as shown in Fig. 9 a～9b, Figure 10.Fig. 9 a, 9b are respectively that resolution is 832 * 480 and 1280 * 720 video sequence PartyScene and the RD curve chart of vidyo1 under the coding method of HM5.0 raw residual and two kinds of methods of Proposed B.Can find out that in Fig. 9 a and Fig. 9 b, two curves almost completely overlap, the coding quality of the method based on Bayes decision theory in this explanation the present invention is almost identical with original method in HM5.0.Figure 10 is the result that HEVC raw residual coding and two kinds of methods of Proposed B are tested the video sequence of different resolution in the situation that of different Q P, just the same in leading indicator and Fig. 8.Can find out that in the situation that maintenance coding quality is constant (coding bit rate on average increases by 0.3%, and the PSNR of brightness on average reduces 0.02dB), the optimization method residual coding time average in the present invention has reduced approximately 40%.
By above-mentioned each chart, can be found out, the present invention is by carrying out hypothesis testing to residual block coefficient, and utilize bayesian theory to end in advance judgement to residual reference block TU, thereby reduce the scramble time of dct transform and quantizing process, at coding quality, lose in negligible situation, significantly reduced the time of HEVC residual coding.
Claims (5)
1. utilize residual error coefficient distribution characteristics and Bayes' theorem to optimize a method for HEVC residual coding, it is characterized in that:
(1) input video sequence: each two field picture that reads video sequence according to the order of setting in control documents;
In frame (2)/inter prediction: the brightness of every two field picture and chromatic value are carried out to infra-frame prediction and inter prediction, thereby try to achieve residual error coefficient;
(3) TU ends to judge in advance: according to residual error coefficient, residual block TU (Transform Unit) (being 32 * 32 to the maximum) is carried out texture judgement or carries out Bayesian model judgement, thereby determine whether to end in advance block encoding;
(4) dct transform and quantification: monoblock TU is carried out to dct transform and quantification, thereby try to achieve quantization parameter;
(5) entropy coding: quantization parameter is carried out to entropy coding, the last formal output with bit stream.
2. the method for utilizing residual error coefficient distribution characteristics and Bayes' theorem to optimize HEVC residual coding according to claim 1, the concrete steps of it is characterized in that in described step (2) frame/inter prediction are as follows:
(2-1) infra-frame prediction be utilize the encoding block left side and top coded reference pixel thereby current block is predicted and is eliminated video image redundancy spatially; Thereby inter prediction carries out to it time redundancy that Motion estimation and compensation is eliminated video sequence by reference to the front and back frame of encoding block place frame; HEVC according to the criterion of rate distortion optimum in a plurality of frames and select in optimum frame and inter-frame forecast mode interframe candidate pattern; Rate distortion costs function is:
Wherein
for rate distortion costs value,
for predicted distortion value,
for the bit number of exporting under different predictive modes,
for LaGrange parameter;
(2-2) H.264 different from video encoding standard before, for improving code efficiency, HEVC adopts larger encoding block (64 * 64), the coded system of circular recursion, intra prediction mode is increased to 35 kinds, and these improve and when improving code efficiency, also make to become more complicated with inter prediction in frame;
If (2-3) input is I frame, only it is carried out infra-frame prediction and carries out rate-distortion optimization; If P frame or B frame, first carry out Skip model prediction and inter prediction to it, and then carry out infra-frame prediction.Finally by rate-distortion optimization standard, from three kinds of patterns, select optimum predictive mode.
3. the method for utilizing residual error coefficient distribution characteristics and Bayes' theorem to optimize HEVC residual coding according to claim 1, is characterized in that described step (3) TU ends the concrete steps judged in advance as follows:
1. the hypothesis test based on Gaussian Profile
From pertinent literature research, it is zero Gaussian Profile that residual error coefficient is obeyed desired value, meets following formula:
（2）
Wherein
represent residual error coefficient,
for average,
for variance,
represent Gaussian Profile;
For judging whether monoblock TU is applicable to being divided into four sub-TU and monoblock TU can be divided into two kinds of dividing mode, as shown in Figure 3.If two kinds of dividing mode all meet, can think that TU is applicable to being divided into four sub-TU.If the dividing mode encoding efficiency in employing Fig. 3 is better, residual error coefficient should Gaussian distributed, owing to obeying the desired value of two parts of same distribution
should there is no difference.Therefore, can judge by hypothesis testing and the prediction effect of this pattern obtain according to the knowledge of probability theory:
（3）
In above formula
with
the number of residual error coefficient in A and two sub-blocks of B in difference presentation graphs 3,
be distributed as Xue Shengshi and distribute,
the degree of freedom of this distribution,
,
the average of two parts pixel intensity in every kind of dividing mode,
,
two parts mathematical expectations separately, in above formula
as follows:
In above formula
with
the value of residual error coefficient in A and two sub-blocks of B in difference presentation graphs 3; Obviously, 32 * 32 macro blocks are divided into two equal parts, the number of pixels of two parts should equate, that is:
, according to hypothesis testing, should meet, get significance level=0.05 simultaneously, looking into t distribution table can obtain
, obtain as drawn a conclusion:
Above formula is equivalent to:
Above formula is to calculate for 32 * 32 TU; In HEVC residual coding, residual block TU adopts and encoding block CU(Coding Unit) similar quadtree coding structure, except 32 * 32, also to carry out traversal formula coding to its sub-block 16 * 16,8 * 8 and 4 * 4, then from all patterns, select optimization model, we can carry out 16 * 16 and 8 * 8 TU similarly processing with 32 * 32 for this reason, thereby obtain
Above-mentioned three formulas are respectively to calculate the Rule of judgment of 32 * 32,16 * 16 and 8 * 8.If meet formula (6), can think not significantly difference of this kind of division in 32 * 32TU, if meet formula (7), (8), can think 16 * 16, the not significantly difference of this kind of division in 8 * 8TU.Residual error coefficient is relevant with coding mode, and simple for what calculate, the treatment step based on above analysis is as follows:
1) first, adopt the monoblock TU to 32 * 32 to carry out residual coding;
2) when two kinds of dividing mode based on residual error coefficient branch characteristics algorithm all do not have significantly difference, redirect 3); Otherwise the TU degree of depth adds 1, and return to 1);
3) monoblock TU is carried out to piecemeal residual coding.
2. the decision algorithm based on bayesian theory
The TU of a certain size carries out block encoding and non-block encoding is two events of opposition completely, by this event definition of TU block encoding, is
, and TU does not carry out block encoding, be not defined as
; TU to measure feature
by the accuracy of helping improve classification, wherein
deng representing piecemeal some characteristic values in close relations whether with TU;
that TU characteristic vector is divided into while being F
the posterior probability of class; In residual coding process, if make the decision making mistake, should block encoding in fact do not have a block encoding, and original block encoding that should piecemeal, this will cause rate distortion loss; We are should piecemeal and actually do not have the loss that piecemeal causes to be labeled as
, and original should piecemeal in fact piecemeal the loss that causes be labeled as
; Between them, there is following relation:
Wherein,
with
it is the rate distortion costs producing when TU carries out block encoding and non-block encoding; As shown from the above formula when TU makes selecting properly, should block encoding in fact also piecemeal or not piecemeal in fact really do not have piecemeal yet, will can not cause any rate distortion loss, therefore
with
be 0.Based on above analysis, can obtain artificial situation
time Bayes risk cost
:
（11）
When
<
time, select
the cost causing is less, should end TU to carry out block encoding; And work as
>
time, select
the cost causing is less, should to TU, carry out block encoding according to original algorithm; In above formula:
Wherein
while representing that TU carries out Splite and None-Splite coding, the probability density function of its characteristic vector F,
it is the priori probability density function of situation Splite and None-Splite; By formula (11), (12) and (13), can be obtained that whether TU is carried out to the judgement formula of block encoding is as follows:
For the probability distribution effectively doping, herein using the mean square deviation MAD of residual block coefficient and and the predicated error of current TU piece as the key element to measure feature F; The non-parametric density estimation technique proposing according to people such as D.Chai is tried to achieve conditional probability density function
and be placed in a question blank; For reducing statistical work amount, two characteristic vector key elements can be quantized into 10 deciles, whole like this characteristic vector F will be divided into 100 scales.And in (14) formula in the decision threshold of the inequality left side
,
,
with
resolution, quantization parameter QP(Quantization Parameter with video) size and the degree of depth of TU all relevant; Therefore, in advance the video sequence of one group of different resolution is added up under different QP, thereby obtained the decision threshold under difference minute variability, different Q P and the different TU degree of depth, and these threshold values are placed in another one question blank.
In sum, the TU type decision method based on bayesian theory comprises following step:
1) decision threshold on inequality the right in (14) formula is carried out to initialization, thus obtain can be to video resolution and
The decision threshold of QP adaptive change;
2) to a TU piece from its root node (32 * 32), monoblock TU is carried out to residual coding, obtain the MAD of encoding error and residual error coefficient;
3) by the MAD of monoblock encoding error and residual error coefficient table look-up find out TU to measure feature F probability-distribution function
with
; By (14) formula, judge, if inequality meets the demands, jump to (4); Otherwise TU is divided into four identical sub-TU, depth D epth adds 1, forwards (2) to;
4) by the order of lining by line scan, next TU is processed equally.
4. the method for utilizing residual error coefficient distribution characteristics and Bayes' theorem to optimize HEVC residual coding according to claim 1, is characterized in that: the concrete steps of described step (4) dct transform and quantification are as follows:
When 1. HEVC carries out integer transform and quantizes, the full-size of TU is 32 * 32, and minimum dimension is 4 * 4, is similar to the quad-tree structure of CU;
2. in different CU, PU, frame and inter prediction, TU has different available types.
5. the method for utilizing residual error coefficient distribution characteristics and Bayes' theorem to optimize HEVC residual coding according to claim 1, is characterized in that: the concrete steps of described step (4) entropy coding are as follows:
1. quantization parameter is carried out to variable entropy coding (VLC) or arithmetic coding (CABAC), thereby the symbol redundancy of remove quantization coefficient realizes the further compression to video sequence;
2. through entropy coded data finally with the formal output of bit stream; The adaptively changing that can realize bit rate by relevant code check control technology, this has improved the network friendliness of HEVC encoder greatly.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201310480055.2A CN103546749B (en) | 2013-10-14 | 2013-10-14 | Method for optimizing HEVC (high efficiency video coding) residual coding by using residual coefficient distribution features and bayes theorem |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201310480055.2A CN103546749B (en) | 2013-10-14 | 2013-10-14 | Method for optimizing HEVC (high efficiency video coding) residual coding by using residual coefficient distribution features and bayes theorem |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN103546749A true CN103546749A (en) | 2014-01-29 |
| CN103546749B CN103546749B (en) | 2017-05-10 |
Family
ID=49969744
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201310480055.2A Active CN103546749B (en) | 2013-10-14 | 2013-10-14 | Method for optimizing HEVC (high efficiency video coding) residual coding by using residual coefficient distribution features and bayes theorem |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN103546749B (en) |
Cited By (14)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103873866A (en) * | 2014-03-18 | 2014-06-18 | 天津大学 | GGoP level distortion recursion method of multi-viewpoint stereoscopic video in packet loss network |
| CN104038760A (en) * | 2014-06-13 | 2014-09-10 | 南京理工大学 | Intra-frame wedge segmentation mode selection method and system for 3D (three-dimensional) video depth image |
| CN104079937A (en) * | 2014-07-16 | 2014-10-01 | 山东大学 | H.264-to-HEVC fast inter-frame transcoding method and device based on motion vector analysis |
| CN105681797A (en) * | 2016-01-12 | 2016-06-15 | 四川大学 | Prediction residual based DVC-HEVC (Distributed Video Coding-High Efficiency Video Coding) video transcoding method |
| WO2018076827A1 (en) * | 2016-10-26 | 2018-05-03 | 北京大学深圳研究生院 | Code rate estimation method for intra-frame coding in video coding |
| CN109168002A (en) * | 2018-07-26 | 2019-01-08 | 西安电子科技大学 | Vision signal measurement field estimation method based on compressed sensing and convolutional neural networks |
| CN109743575A (en) * | 2018-12-05 | 2019-05-10 | 四川大学 | A kind of DVC-HEVC video transcoding method based on naive Bayesian |
| CN110234011A (en) * | 2019-04-19 | 2019-09-13 | 网宿科技股份有限公司 | A kind of video-frequency compression method and system |
| CN110430011A (en) * | 2019-07-09 | 2019-11-08 | 武汉大学 | The BATS code encoding method of rule-based variable node degree distribution |
| CN110913215A (en) * | 2019-12-03 | 2020-03-24 | 北京数码视讯软件技术发展有限公司 | Method and device for selecting prediction mode and readable storage medium |
| WO2020244022A1 (en) * | 2019-06-03 | 2020-12-10 | Tcl华星光电技术有限公司 | Compensation table compression method, apparatus for manufacturing display, and memory |
| CN113301347A (en) * | 2021-05-08 | 2021-08-24 | 广东工业大学 | Optimization method of HEVC high-definition video coding |
| CN115278235A (en) * | 2022-09-27 | 2022-11-01 | 北京达佳互联信息技术有限公司 | Video coding method and device, electronic equipment and storage medium |
| CN117294861A (en) * | 2023-11-24 | 2023-12-26 | 淘宝（中国）软件有限公司 | Coding block dividing method based on inter-frame prediction and coder |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1674677A (en) * | 2005-04-21 | 2005-09-28 | 上海大学 | Improved FGS coding method and coder/decoder thereof |
| CN101106695A (en) * | 2007-03-29 | 2008-01-16 | 上海大学 | Real time fine flexible coding method based on H.264 |
| EP2063645B1 (en) * | 2007-11-20 | 2011-02-16 | Kabushiki Kaisha Toshiba | Coding apparatus, coding method, and coding system |
-
2013
- 2013-10-14 CN CN201310480055.2A patent/CN103546749B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1674677A (en) * | 2005-04-21 | 2005-09-28 | 上海大学 | Improved FGS coding method and coder/decoder thereof |
| CN101106695A (en) * | 2007-03-29 | 2008-01-16 | 上海大学 | Real time fine flexible coding method based on H.264 |
| EP2063645B1 (en) * | 2007-11-20 | 2011-02-16 | Kabushiki Kaisha Toshiba | Coding apparatus, coding method, and coding system |
Cited By (23)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103873866A (en) * | 2014-03-18 | 2014-06-18 | 天津大学 | GGoP level distortion recursion method of multi-viewpoint stereoscopic video in packet loss network |
| CN104038760A (en) * | 2014-06-13 | 2014-09-10 | 南京理工大学 | Intra-frame wedge segmentation mode selection method and system for 3D (three-dimensional) video depth image |
| CN104038760B (en) * | 2014-06-13 | 2017-03-15 | 南京理工大学 | A kind of wedge shape Fractionation regimen system of selection of 3D video depths image frame in and system |
| CN104079937A (en) * | 2014-07-16 | 2014-10-01 | 山东大学 | H.264-to-HEVC fast inter-frame transcoding method and device based on motion vector analysis |
| CN104079937B (en) * | 2014-07-16 | 2017-06-23 | 山东大学 | It is a kind of based on motion vector analysis by H.264 to the quick interframe code-transferring method and transcoding device of HEVC |
| CN105681797A (en) * | 2016-01-12 | 2016-06-15 | 四川大学 | Prediction residual based DVC-HEVC (Distributed Video Coding-High Efficiency Video Coding) video transcoding method |
| CN105681797B (en) * | 2016-01-12 | 2018-09-04 | 四川大学 | A kind of DVC-HEVC video transcoding methods based on prediction residual |
| US10917646B2 (en) | 2016-10-26 | 2021-02-09 | Peking University Shenzhen Graduate School | Intra code-rate predicting with rate distortion optimization method in video coding |
| WO2018076827A1 (en) * | 2016-10-26 | 2018-05-03 | 北京大学深圳研究生院 | Code rate estimation method for intra-frame coding in video coding |
| CN109168002A (en) * | 2018-07-26 | 2019-01-08 | 西安电子科技大学 | Vision signal measurement field estimation method based on compressed sensing and convolutional neural networks |
| CN109743575A (en) * | 2018-12-05 | 2019-05-10 | 四川大学 | A kind of DVC-HEVC video transcoding method based on naive Bayesian |
| CN110234011A (en) * | 2019-04-19 | 2019-09-13 | 网宿科技股份有限公司 | A kind of video-frequency compression method and system |
| CN110234011B (en) * | 2019-04-19 | 2021-09-24 | 网宿科技股份有限公司 | Video compression method and system |
| WO2020244022A1 (en) * | 2019-06-03 | 2020-12-10 | Tcl华星光电技术有限公司 | Compensation table compression method, apparatus for manufacturing display, and memory |
| CN110430011A (en) * | 2019-07-09 | 2019-11-08 | 武汉大学 | The BATS code encoding method of rule-based variable node degree distribution |
| CN110913215A (en) * | 2019-12-03 | 2020-03-24 | 北京数码视讯软件技术发展有限公司 | Method and device for selecting prediction mode and readable storage medium |
| CN110913215B (en) * | 2019-12-03 | 2022-04-12 | 北京数码视讯软件技术发展有限公司 | Method and device for selecting prediction mode and readable storage medium |
| CN113301347A (en) * | 2021-05-08 | 2021-08-24 | 广东工业大学 | Optimization method of HEVC high-definition video coding |
| CN113301347B (en) * | 2021-05-08 | 2023-05-05 | 广东工业大学 | HEVC high definition video coding optimization method |
| CN115278235A (en) * | 2022-09-27 | 2022-11-01 | 北京达佳互联信息技术有限公司 | Video coding method and device, electronic equipment and storage medium |
| CN115278235B (en) * | 2022-09-27 | 2023-01-20 | 北京达佳互联信息技术有限公司 | Video coding method and device, electronic equipment and storage medium |
| CN117294861A (en) * | 2023-11-24 | 2023-12-26 | 淘宝（中国）软件有限公司 | Coding block dividing method based on inter-frame prediction and coder |
| CN117294861B (en) * | 2023-11-24 | 2024-03-22 | 淘宝（中国）软件有限公司 | Coding block dividing method based on inter-frame prediction and coder |
Also Published As
| Publication number | Publication date |
|---|---|
| CN103546749B (en) | 2017-05-10 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN103546749A (en) | Method for optimizing HEVC (high efficiency video coding) residual coding by using residual coefficient distribution features and bayes theorem | |
| CN105049850B (en) | HEVC bit rate control methods based on area-of-interest | |
| CN100488254C (en) | Entropy coding method and decoding method based on text | |
| CN102065298B (en) | High-performance macroblock coding implementation method | |
| CN106937116B (en) | Low-complexity video coding method based on random training set adaptive learning | |
| CN106170092A (en) | Fast encoding method for lossless coding | |
| CN108989802B (en) | HEVC video stream quality estimation method and system by utilizing inter-frame relation | |
| CN104067617A (en) | Method and apparatus for encoding/decoding image | |
| US10554965B2 (en) | Motion-compensated partitioning | |
| CN112738511B (en) | Fast mode decision method and device combined with video analysis | |
| CN103581647A (en) | Depth map sequence fractal coding method based on motion vectors of color video | |
| CN104853191B (en) | A kind of HEVC fast encoding method | |
| CN106791848B (en) | Two-Pass code rate control method based on HEVC | |
| CN104038764A (en) | H.264-to-H.265 video transcoding method and transcoder | |
| CN103546758A (en) | Rapid depth map sequence interframe mode selection fractal coding method | |
| CN106888379B (en) | Applied to the interframe fast video code-transferring method for H.264 arriving HEVC | |
| CN104601992B (en) | SKIP mode quick selecting methods based on Bayesian Smallest Risk decision | |
| CN106412611B (en) | A kind of complexity control method of efficient video coding | |
| CN103442229B (en) | The bit rate estimation method of SAO mode adjudgings suitable for the encoder of HEVC standard | |
| CN103384327A (en) | AVS fast mode selection algorithm based on adaptive threshold | |
| CN104333755B (en) | The CU based on SKIP/Merge RD Cost of B frames shifts to an earlier date terminating method in HEVC | |
| CN103747272B (en) | Fast transform approach for the remaining quaternary tree coding of HEVC | |
| CN104702959A (en) | Intra-frame prediction method and system of video coding | |
| CN101179729A (en) | Interframe mode statistical classification based H.264 macroblock mode selecting method | |
| CN106060555A (en) | Encoder based on multi-core processor |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |