CN104052994A - Method for HEVC intra-frame prediction mode quick decision based on grading self-adaption - Google Patents
Method for HEVC intra-frame prediction mode quick decision based on grading self-adaption Download PDFInfo
- Publication number
- CN104052994A CN104052994A CN201410159107.0A CN201410159107A CN104052994A CN 104052994 A CN104052994 A CN 104052994A CN 201410159107 A CN201410159107 A CN 201410159107A CN 104052994 A CN104052994 A CN 104052994A
- Authority
- CN
- China
- Prior art keywords
- mode
- hcost
- pattern
- cost
- rmd
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Landscapes
- Compression Or Coding Systems Of Tv Signals (AREA)
Abstract
The invention is suitable for the technical field of video coding and decoding and provides a method for intra-frame prediction mode quick decision based on grading self-adaption. The method comprises the steps of adopting a grading calculation and step-to-step elaboration mode screening method, first selecting 12 candidate modes to perform prediction direction initial judgment, and then performing selection of prediction modes again according to an initial judging result; adopting a cost self-adaption method to eliminate part of candidate modes low in possibility in advance according to HCost of an RMD roughing mode, so as to achieve the goal of reducing the intra-frame prediction mode number and calculation complexity; and finally adding an MPM mode to a candidate prediction mode subset, so that coding efficiency and quality can be guaranteed. The method mainly solves the problem that existing HEVC intra-frame prediction coding is high in calculation complexity. The method for intra-frame prediction mode quick decision is easy to achieve, coding time can be reduced by 36.521% on average under the conditions that the BD-PSNR average loss is 0.041dB, and the coding speed is remarkably improved.
Description
Technical field
The invention belongs to video coding and decoding technology field, be specifically related to a kind of classification adaptive intra prediction mode high-speed decision method and device for efficient video coding (High Efficiency video coding, HEVC).
Background technology
For meeting the new demand of HD video application, in April, 2010, JCT-VC starts to formulate new video encoding and decoding standard HEVC of future generation (High Efficiency Video Coding), is also called H.265.
With respect to previous generation standard H.264/AVC, HEVC has had significant progress aspect predictive coding, has improved the code efficiency of approximately 1 times.The HEVC more recent application such as generation 4K resolution HDTV (High-Definition Television), 3D video, mobile video, streaming media video that mainly face down.
At present, the concrete document of each discussion of JCT-VC tissue can obtain from http://wftp3.itu.int, and these technical schemes likely can progressively join in HEVC standard.
Aspect infra-frame prediction (intra predication), 35 kinds of intra prediction modes of the highest support of HEVC, wherein prediction direction is increased to 33 kinds, has improved precision of prediction.In addition, on the basis of the equal binarization mode of DC, increase Planar smoothing prediction pattern, improved the precision of prediction to gradual change type smooth grain district.Fig. 1 shows 35 kinds of intra prediction modes of HEVC.
Owing to having used nearly 35 kinds of intra prediction modes, if directly for 35 kinds of intra prediction mode utilization rate aberration optimizing (RateDistortion optimization, RDO) technology judges optimal prediction modes, and the operand of encoder will be too huge.Therefore, how to reduce the computation complexity of HEVC Video coding, become the current study hotspot of HEVC.
The jani Lainema of the jung-Hye MinDeng He Nokia of Samsung etc. are by (the Angular Intra Prediction of angle infra-frame prediction in motion JCTVC-A119, AIP) any direction infra-frame prediction (Arbitrary Directional Intra in technology and motion JCTVC-A124, ADI) technology combination, called after associating direction prediction (Unification of the Directional Intra prediction, UDI).
In motion JCTVC-C207, the Yinji Piao of Samsung etc. is optimized UDI technology, has simplified rate-distortion model, proposes a kind of coarse mode decision-making (Rough Mode Decision, RMD) method, obtains the candidate pattern subset of simplifying.First, calculate HCost=HSATD
mode+ λ bits
modecost, calculates residual signals Hadamard conversion cost and mode bit and counts sum, carries out rough model selection cost and calculates.Filter out the mode subset of roughly selecting that front N kind predictive mode by Least-cost forms; Then this N kind predictive mode of roughly selecting in mode subset is carried out to RDO calculating, RDCost
mode=∑ Diff (i, j)
2+ λ bits
mode, the pattern of RDCost Least-cost is optimal frames inner estimation mode.This algorithm has reduced the quantity of candidate's predictive mode, has reduced the computation complexity of Intra prediction mode selection.
In motion JCTVC-C218, the people such as the jungsun Kim of LG company have simplified the redundancy process that CU is divided into PU, proposition is to predicting unit (Prediction Unit, PU) do not adopt quad-tree partition prediction, and forbid rotation transformation (Rotation Transform, ROT) with combine infra-frame prediction (Combined intra prediction, CIP) technology, reduce by approximately 35% scramble time.
In motion JCTVC-D283, simplified UDI algorithm, and added most possible pattern (most probable mode, MPM) technology.At RMD, roughly select and obtain after N roughly selects mode subset, application MPM algorithm extracts most possible pattern and adds and roughly select mode subset.
After HM7.0, for prediction more accurately, the PU of each size is used 35 kinds of all intra prediction modes to carry out RMD pattern and roughly selects and select.Filter out N candidate pattern of Least-cost, and this N candidate pattern is carried out to ascending sort according to HCost cost size.Wherein, for size, be 4 * 4 and 8 * 8 prediction piece, N=8, is 16 * 16 and 32 * 32 prediction piece for size, N=3.
Although the loss of the coding efficiency of above several fast algorithms is minimum, the amplitude that the scramble time is saved is also little.Especially for the PU piece of 4 * 4 and 8 * 8 sizes, it is more that RMD roughly selects the candidate pattern quantity obtaining, and also needs further investigation.
The application is main to be realized with reference to following technical literature.
[1]J.H.Min，“Unification?of?the?directional?intra?prediction?methods?in?TMuC，”JCTVC-B100，July,2010.
[2]K.McCann?et.al.,“Samsung's?response?to?the?call?for?proposals?on?video?compression?technology,”JCTVC-A124,April,2010.
[3]K.Ugur,K.R.Andersson?and?A.Fuldseth,“Description?of?video?coding?technology?proposal?by?Tandberg,Nokia,Ericsson,”JCTVC-A119,April,2010.
[4]Y.Piao,J.H.Min,J.Chen,“Encoder?improvement?of?unified?intra?Prediction,”JCTVC-C207,Oct.,2010.
[5]L.Zhao,et.al.,“Further?encoder?improvement?of?intra?mode?decision,”JCTVC-D283,Jan.,2011.
[6]J.Kim,Y.Jeon,B.Jeon,“Encoding?complexity?reduction?for?intra?prediction?by?disabling?N×N?partition,”JCTVC-C218,Oct.,2010.
[7]B.BroSS,W.J.Han,J.R.Ohm,G.J.Sullivan,T.Wiegand,“WD4:Working?Draft?4?of?High-Efficiency?Video?Coding,”JCTVC-F803,July,2011.
[8]K.McCann,5.Sekiguei,B.Bross,W.J.Han,“HM4:HEVC?Test?Model4Encoder?Description,”JCTVC-F802,July,2011.
[9]Y.H.Tan,C.Yeo,H.L.Tan,Z.Li,“RQT?depth?selection,”JCTVC-E104,Mar.,2011.
[10]K.Panusopone,K.Chono,Y.H.Tan,M.Zhou,“Evaluation?of?RQT?in?HM?and?related?TUrepresentation,”JCTVC-E365,Mar.,2011.
[11]B.Bross,“Common?test?conditions?and?software?reference?configurations,”JCTVC-F900,July,2011.
Summary of the invention:
The invention provides the adaptive HEVC intra prediction mode of a kind of classification high-speed decision method and device, the method is the process of intra prediction mode decision-making of carrying out for each the prediction piece (PU) in present frame, and emphasis is for pattern, to roughly select the decision process of selecting (RMD).
The object of the invention is to: for the large deficiency of Intra prediction mode selection process computation amount in above-mentioned HEVC, propose a kind of classification self-adaption intra-frame prediction pattern high-speed decision method for HEVC, to reduce the computation complexity of intra prediction mode judgement.
Realizing the object of the invention technical thought is: on the basis of HEVC, further propose classification coarse mode Decision Method (Classification Rough Mode Decision, CRMD), the general orientation of first searching optimum prediction direction place is interval, then the predictive mode in general orientation interval is carried out to fine search, in the hope of more accurate predictive mode; And adopt the in advance exclusive method (Early Rule Out, ERO) of cost adaptive, according to HCost cost, shift to an earlier date the candidate pattern that exclusive segment possibility is less, to improve the speed of HEVC intraframe coding; Finally add MPM pattern to roughly selecting in mode subset, thereby ensured code efficiency and quality.
The present invention is directed to each PU piece in HEVC encoding scheme and carry out Intra prediction mode selection part, therefore keep other parts of HEVC encoding scheme constant.
First aspect, for achieving the above object, the concrete technical scheme that the present invention adopts is:
Step 1) adopt 12 one-level candidate pattern as shown in Figure 2 to carry out RMD and roughly select calculating, determine that the general orientation at optimum prediction direction place is interval.Planar pattern in one-level candidate pattern is mainly considered the decision-making of texture smooth region, and all the other 11 directional prediction modes adopt the mode at even angle interval to select.
Step 2) after step 1 is calculated, select 2 predictive modes (minimum cost pattern Mode_HCost0, inferior little cost pattern Mode_HCost1) of HCost Least-cost, and detect in these two better predictive modes whether have Planar pattern.If there is Planar pattern, proceed to step 3; Otherwise proceed to step 4.
Step 3) increase by 3 secondary candidates and carry out RMD and roughly select calculating, be respectively the adjacent predictive mode in 2 of left and right of DC pattern, Mode_HCost0, calculate candidate pattern scope accurately; And according to HCost cost, all candidate pattern of carrying out RMD calculating are arranged by ascending order: proceed to step 5.
Step 4) increase by 4 secondary candidates and carry out RMD and roughly select calculating, be respectively the adjacent predictive mode in 2 of left and right of Mode_HCost0 and the adjacent predictive mode in 2 of left and right of Mode_HCost1, calculate that prediction direction is interval accurately; And according to HCost cost, all candidate pattern of carrying out RMD calculating are arranged by ascending order.
Step 5) determine for the first time RMD predictive mode the selection result, filter out the front N kind predictive mode of HCost Least-cost as roughly selecting mode subset.PU size for 16 * 16,32 * 32 and 64 * 64, N=3; PU size for 4 * 4 and 8 * 8, N=2+ secondary candidate pattern number.
Step 6) carry out the eliminating in advance (ERO) of HCost cost adaptive.The HCost cost size of each predictive mode in mode subset is roughly selected in investigation, if its cost HCost
mode> α * HCost
0, excluded and roughly selected mode subset.According to coding experience α=1.2, HCost
0finger RMD pattern is roughly selected the minimum cost obtaining.If cost difference is large, illustrate that the grain direction of current PU piece is obvious, can get rid of the candidate pattern that a part of possibility is less accordingly.After ERO gets rid of in advance, the candidate pattern number of roughly selecting in the inventive method in mode subset will dynamically be reduced to 1-7.
Whether step 7) detect most possible pattern MPM is included in and roughly selects in mode subset.If do not comprise, joined and roughly selected mode subset, final definite mode subset of roughly selecting.
Step 8) last, to roughly selecting the predictive mode calculation rate distortion cost RDCost in mode subset, the pattern of Least-cost is final optimal prediction modes.
Second aspect, provides a kind of intra prediction mode decision making device for HEVC efficient video coding, and described intra prediction mode decision making device is to carry out predictive mode decision-making for each the prediction piece (PU) in present frame; It is characterized in that, described device is configured to:
(1) adopt 12 one-level candidate pattern as shown in Figure 2 to carry out RMD and roughly select calculating, select 2 better predictive modes (minimum cost pattern Mode_HCost0 of Least-cost, inferior little cost pattern Mode_HCost1), determine the general orientation interval at optimum prediction direction place.
(2) while there is Planar pattern in these two better predictive modes (minimum cost pattern Mode_HCost0 and time little cost pattern Mode_HCost1),
Increase by 3 secondary candidates and carry out RMD and roughly select calculating, the adjacent predictive mode in 2 of left and right of DC pattern, Mode_HCost0, calculates candidate pattern scope accurately.
While there is not Planar pattern in these two better predictive modes (minimum cost pattern Mode_HCost0 and time little cost pattern Mode_HCost1),
Increase by 4 secondary candidates and carry out RMD and roughly select calculating, i.e. the adjacent predictive mode in 2 of left and right of the adjacent predictive mode in 2 of left and right of Mode_HCost0 and Mode_HCost1, calculates that prediction direction is interval accurately.
(3) determine for the first time RMD predictive mode the selection result, obtain the mode subset of roughly selecting that comprises N candidate pattern.PU size for 16 * 16,32 * 32 and 64 * 64, N=3; PU size for 4 * 4 and 8 * 8, N=2+ secondary candidate pattern number.
(4) carry out the eliminating in advance (ERO) of HCost cost adaptive.The cost size of each predictive mode in mode subset is roughly selected in investigation, if its cost HCost
mode> α * HCost
0, excluded and roughly selected mode subset.According to coding experience α=1.2, HCost
0finger RMD pattern is roughly selected the minimum cost obtaining.
(5) use most possible pattern (MPM) technology, wherein, described in two most possible patterns are added, roughly select mode subset, final definite mode subset of roughly selecting.
(6), to described each pattern of roughly selecting in mode subset, adopt rate-distortion optimization (RDO) method to obtain described optimal frames inner estimation mode.
The video encoder that also discloses the device corresponding with said method and be used for carrying out said method.
The present invention compared with prior art has the following advantages:
First, the present invention proposes on HEVC basis 35 intra prediction modes to take that classification is calculated, meticulous coarse mode Decision Method (CRMD) progressively, one-level candidate pattern contains 12 patterns, secondary candidate pattern contains 3-4 pattern, altogether need the pattern of carrying out HCost calculating to have 15～16, no more than all 35 intra prediction mode quantity half, saved the RMD amount of calculation of half.
The second, the present invention proposes the exclusive method (ERO) in advance of HCost cost adaptive, can shift to an earlier date the candidate pattern that exclusive segment possibility is less.After ERO gets rid of in advance, to the PU piece of 16 * 16 above sizes, the candidate pattern of this algorithm will dynamically be reduced to 1～3; PU piece to 8 * 8 following sizes, candidate pattern will dynamically be reduced to 1～6, reduce complicated RDCost calculation times, improve arithmetic speed.
The 3rd, the present invention proposes to roughly select CRMD and adaptive pattern is got rid of after ERO in advance at hierarchical pattern, add again most possible pattern MPM, can significantly reduce candidate pattern quantity, can optimal prediction modes be retained in and be roughly selected in mode subset with high probability again, thereby keep Image Coding performance substantially constant.
The 4th, implementation method of the present invention is simple, and this programme only needs to get final product by simple packet, sequence and compare operation the selection course of achieve frame inner estimation mode.Method of the present invention, the in the situation that of BD-PSNR average loss 0.041dB, on average can reduce for 36.521% scramble time, has significantly improved coding rate.
Accompanying drawing explanation
Fig. 1 shows 35 kinds of intra prediction modes that use in HEVC.
Fig. 2 shows 12 one-level candidate pattern of the present invention.
Fig. 3 shows the probability distribution graph that each RMD pattern of roughly selecting becomes optimization model.
Fig. 4 shows intra prediction mode decision flow diagram according to an embodiment of the invention.
Embodiment
Below in conjunction with drawings and Examples, the present invention is further elaborated.Should be appreciated that specific embodiment described herein, only in order to explain the present invention, is not intended to limit the present invention.
At HM7.0 (HEVC test model7.0) afterwards, for prediction more accurately, the prediction piece (PU) of each size has 35 kinds of predictive modes.As shown in Figure 1,35 kinds of intra prediction modes comprise 33 kinds of directional prediction modes, and DC (average) pattern and Planar (smoothly) pattern.
After HM7.0, the judgement of HEVC intra prediction mode has three steps.First step pattern is roughly selected and is selected (RMD), and 35 all candidate pattern are carried out HCost calculating, from filter out the front N kind predictive mode of Least-cost, roughly select mode subset.N=8, the PU size for 4 * 4 and 8 * 8; N=3, for other PU size.Second step MPM technology, whether the most possible pattern MPM that detects present encoding piece is included in and roughly selects in subset.If do not exist, added and roughly selected subset.The 3rd step rate distortion judgement (RDO), carries out RDCost calculating by each pattern of roughly selecting in mode subset, and the pattern of Least-cost is optimal prediction modes.
First step RMD technology
Particularly, in RMD, with following formula, select N candidate pattern of Least-cost, and this N candidate pattern is carried out to ascending sort according to HCost cost size.Wherein, for size, be 4 * 4 and 8 * 8 prediction piece, N=8; For size, be 16 * 16 and 32 * 32 prediction piece, N=3.
HCost＝HSATD
mode+λ·bits
m。de，?(1)
Wherein, HCost is cost, and mode is all intra prediction modes of 35 kinds of being up to of HEVC regulation, HSATD represent residual signals Hadamard transform coefficients definitely and, bits represents the consumption figures of mode bit number, λ represents Lagrangian constant.
Second step MPM technology
Particularly, after using RMD technology to obtain to roughly select mode subset, the optimization model of the left side block of current prediction piece and upper block, directly as two most possible patterns (MPM), and is judged to whether MPM pattern is included in and roughly selects in mode subset; If do not comprised, added and roughly selected mode subset.
The 3rd step RDO technology
RDO technology is the technology that field of video encoding generally adopts, and it determines optimal prediction modes based on residual signals quadratic sum (Sum of Square Error, SSE) conventionally.In order not obscure key, it is not launched to describe herein.
Our experiment observation and data analysis:
Table 1 has shown that RMD roughly selects the probability statistics that pattern and MPM pattern finally become current PU piece optimal prediction modes.The average probability that the pattern Mode_HCost0 that RMD roughly selects rear Least-cost finally becomes current PU optimal prediction modes has reached 71.9%; And, from Mode_HCost0 to Mode_HCost7, along with candidate pattern is roughly selected the reduction of middle rank at RMD, present the probability trend of decay fast, as shown in Figure 3.
Data are herein results that (encoder_intra_main configuration) carries out the full I frame coding of 50 frame statistics on the reference software HM12.0 of HEVC, and experimental subjects is Class A～E 15 standard test sequences of totally 5 classifications.
The table 1RMD pattern of roughly selecting becomes the probability statistics (QP=32, the full I frame of 50 frame) of current prediction piece (PU) optimal prediction modes
Utilize above-mentioned statistical property, in the adaptive HEVC intra prediction mode of the grouping high-speed decision method proposing in the present invention, by adopting hierarchical pattern to roughly select CRMD and adaptive pattern shifts to an earlier date elimination technique ERO, reduce and roughly select the candidate pattern quantity in mode subset, to reduce the computation complexity of HEVC infra-frame prediction.For technical solutions according to the invention are described, below by specific embodiment, describe.
Embodiment:
Fig. 4, shows the HEVC intra prediction mode high-speed decision flow chart of one embodiment of the invention, for convenience of explanation, only shows the part relevant to the embodiment of the present invention.
The HEVC intra prediction mode high-speed decision method that this example provides comprises:
In step 401, employing 12 one-level candidate pattern are as shown in Figure 2 carried out RMD and are roughly selected calculating, determine that the general orientation at optimum prediction direction place is interval.Particularly, use formula (1) to carry out the HCost cost of calculated candidate pattern, and according to HCost cost, candidate pattern is arranged by ascending order.
In step 402, in two patterns of inspection HCost Least-cost, whether there is Planar pattern.If there is Planar pattern, proceed to step 403; Otherwise proceed to step 404.
In step 403, increase by 3 secondary candidate pattern and carry out RMD and roughly select calculating, be respectively the adjacent predictive mode in 2 of left and right of DC pattern, better pre-measuring angle Mode_HCost0, calculate candidate pattern scope accurately.Particularly, use formula (1) to calculate the HCost cost of 3 newly-increased candidate pattern, and according to HCost cost, all candidate pattern of carrying out RMD calculating are arranged by ascending order.Proceed to step 405.
In step 404, increase by 4 secondary candidate pattern and carry out RMD and roughly select calculating, be respectively the adjacent predictive mode in 2 of left and right of Mode_HCost0 and the adjacent predictive mode in 2 of left and right of Mode_HCost1, calculate candidate pattern scope accurately.Particularly, use formula (1) to calculate the HCost cost of 4 newly-increased candidate pattern, and according to HCost cost, all candidate pattern of carrying out RMD calculating are arranged by ascending order.Proceed to step 405.
In step 405, determine for the first time RMD predictive mode the selection result, filter out the front N kind predictive mode of Least-cost as roughly selecting mode subset.Particularly, the PU size for 16 * 16,32 * 32 and 64 * 64, N=3; PU size for 4 * 4 and 8 * 8, N=2+ secondary candidate pattern number.Turn people's step 406.
In step 406, adopt the exclusive method (ERO) in advance of cost adaptive, the candidate pattern that exclusive segment possibility is less.Particularly, relatively roughly select the HCost cost size of each candidate pattern in mode subset, if its cost HCost
mode> 1.2 * HCost
0, excluded and roughly selected mode subset.HCost
0finger RM) pattern is roughly selected the minimum HCost cost obtaining.After ERO gets rid of in advance, the candidate pattern number that this algorithm is roughly selected in subset will dynamically be reduced to 1-6, can effectively reduce the pattern quantity that need to carry out RDO calculating.Proceed to step 407.
In step 407, use MPM technology, using the optimization model of the left side block of current prediction piece and upper block as two most possible patterns, add and roughly select mode subset.Whether particularly, detect most possible pattern MPM is included in and roughly selects in subset.If do not comprise, joined and roughly selected mode subset.Proceed to step 408.
In step 408, use RDO technology, to roughly selecting the candidate pattern calculation rate distortion cost RDCost in mode subset, the pattern of Least-cost is final optimal prediction modes.
So far completed the decision process of the interior optimal prediction modes of frame of a predicting unit (PU) in HEVC.
After having determined the optimal frames inner estimation mode of current block, encoder can be carried out infra-frame prediction to present frame according to this optimal frames inner estimation mode, and to prediction residual convert, quantification and entropy coding.As mentioned above, the concrete residual coding process of infra-frame prediction does not belong to the emphasis that the present invention pays close attention to, and therefore repeats no more in this article.
Experiment test of the present invention and Data Comparison are all the reference software HM12.0 based on HEVC, and test environment is to set according to the requirement of motion JCTVC-F900.Specific experiment parameter is set to: encoder_intra_main configuration, IntraPeriod=1 (being full I frame coding), front 50 frames of encoding, frame frequency 30fps, MaxCUWidth=64, MaxCUHeight=64, MaxPatitionDepth=4, and CABAC entropy coding.
Because the present invention focuses on infra-prediction techniques, so we adopt full I frame to carry out the encoded test of 50 frames.For more objective, more fully the coding efficiency of described method is compared, we have selected the standard YUV cycle tests of 5 kinds of different sizes such as HEVC Class A～E to carry out experiment test.In cycle tests, Class A is that resolution is 2560 * 1600 super clear sequence, Class B and Class E are that resolution is 1920 * 1080 and 1280 * 720 high definition and SD sequence, and Class C and Class D are respectively 832 * 480 and 416 * 240 WVGA and WQVGA small size sequence.Table 2 and table 3 have reflected the coding efficiency contrast situation of intra-frame prediction method in institute's extracting method of the present invention and HM12.0.
The experiment test comparison of intra-frame prediction method in table 2 institute of the present invention extracting method and HM12.0.
According to the experimental data of table 2, the present invention is based on the adaptive intra prediction mode high-speed decision of classification method and there is good coding efficiency, code check on average increases by 0.164%, and brightness PSNR on average reduces 0.032dB; This loss is very little on the impact of coding RD performance, and human eye is discovered the loss less than image quality substantially.Meanwhile, the cycle tests to different size, different texture and movement characteristic, the inventive method has all obtained roughly the same effect, demonstrates good adaptability, more closing to reality application.
BD-PSNR and the scramble time of table 3 institute of the present invention extracting method, with respect to HM12.0
BD-PSNR represented under same code rate, the difference of the brightness PSNR of institute's extracting method and HM12.0 method herein, computational methods that specifically can list of references VCEG-AE07.From the BD-PSNR experimental data of table 3, can find out, under identical code check, the mean P SNR of institute of the present invention extracting method loss only has 0.041dB, and scramble time average energy shortens 36.521%.
The adaptive HEVC intra prediction mode of classification disclosed in this invention high-speed decision method can realize with software, hardware, firmware etc.
When realizing with hardware, video encoder can or be designed to carry out its combination in any of function described herein by general processor, digital signal processor (DsP), application-specific integrated circuit (ASIC) (ASIC), field programmable gate array (FPGA) or other programmable logic device, discrete gate or transistor logic device, discrete hardware components, realizes or carries out.
When realizing video encoder with hardware circuits such as ASIC, FPGA, it can comprise the various circuit blocks that are configured to carry out various functions.Those skilled in the art can carry out these circuit of Design and implementation in every way according to the various constraintss that are applied in whole system, realize various function disclosed in this invention.
Although aforementioned open file has been discussed exemplary arrangement and/or embodiment, it should be noted that in the situation that do not deviate from the scheme of the description being defined by claims and/or the scope of embodiment, can make many variations and modification at this.
Above-mentioned calculating and operating process are a kind of typical execution modes of the method for the invention, for those skilled in the art, can do the variation of multiple searching algorithm on the basis of above-described embodiment, can realize object of the present invention equally.But this variation obviously should be in the protection range of claims of the present invention.
Claims (8)
1. the adaptive HEVC intra prediction mode of a classification high-speed decision method, wherein, described method is the method for predictive mode decision-making of carrying out for each the prediction piece (PU) in present frame, it is characterized in that, comprises the following steps:
Step 1) adopt 12 one-level candidate pattern at even angle interval to roughly select calculating (RMD), determine that the general orientation at optimum prediction direction place is interval;
Step 2) the HCost cost calculating according to RMD is arranged by ascending order candidate pattern; In two patterns of inspection HCost Least-cost, whether there is Planar pattern; If there is Planar pattern, proceed to step 3; Otherwise proceed to step 4;
Step 3) if there is Planar pattern in two patterns of HCost Least-cost, increase by 3 secondary candidate pattern and carry out RMD and roughly select calculating, be respectively the adjacent predictive mode in 2 of left and right of DC pattern, better pre-measuring angle Mode_HCost 0; And according to HCost cost, all candidate pattern of carrying out RMD calculating are arranged by ascending order; Enter step 5;
Step 4) if there is not Planar pattern in two patterns of HCost Least-cost, increase by 4 secondary candidate pattern and carry out RMD and roughly select calculating, be respectively the adjacent predictive mode in 2 of left and right of Mode_HCost 0 and the adjacent predictive mode in 2 of left and right of Mode_HCost 1; And according to HCost cost, all candidate pattern of carrying out RMD calculating are arranged by ascending order; Enter step 5;
Step 5) determine for the first time RMD predictive mode the selection result, filter out the front N kind predictive mode of HCost Least-cost as roughly selecting mode subset; Particularly, the PU size for 16 * 16,32 * 32 and 64 * 64, N=3; PU size for 4 * 4 and 8 * 8, N=2+ secondary candidate pattern number;
Step 6) adopt the exclusive method in advance of cost adaptive, the candidate pattern that exclusive segment possibility is less; Particularly, relatively roughly select the HCost cost size of each predictive mode in mode subset, if its cost HCost
mode> 1.2 * HCost
0, excluded and roughly selected mode subset; Determine that the RMD after screening roughly selects mode subset;
Step 7) use most possible pattern (MPM) method, using the optimal mode of the left side block of current prediction piece and upper block as two most possible patterns, add and roughly select mode subset;
Step 8) utilization rate distortion (RDO) method is determined optimal prediction modes, and particularly, to the described predictive mode calculation rate distortion cost RDCost roughly selecting in subset, the pattern of Least-cost is final optimal prediction modes.
2. the method for claim 1, is characterized in that, 12 one-level candidate pattern at wherein said even angle interval refer to the intra prediction modes such as pattern 0,3,6,9,12,15,18,21,24,27,30,33 of HEVC regulation.
3. the method for claim 1, it is characterized in that, wherein said secondary candidate pattern is the result of calculation according to step 1 and step 2, carry out Dynamic Selection, mainly comprise the adjacent predictive mode in 2 of left and right of DC pattern, Mode_HCost 0 and the adjacent predictive mode in 2 of left and right of Mode_HCost 1.
4. the method for claim 1, is characterized in that, wherein said in advance exclusive method based on below:
HCost
mode＞1.2×HCost
0
Wherein, HCost
0for RMD pattern is roughly selected the minimum cost obtaining; Owing to carrying out RMD while roughly selecting, will to candidate pattern, by ascending order, arrange according to HCost cost, therefore RMD pattern is roughly selected the minimum cost HCost obtaining
0represent; RMD corresponding thereto roughly selects the minimum cost pattern obtaining and represents with Mode_HCost 0.
5. the HCost cost that the method for claim 1, wherein described RMD roughly selects calculating based on below:
HCost＝HSATD
mode+λ·bits
mode
Wherein, mode is one of 35 kinds of intra prediction modes of HEVC regulation, HSATD represent residual signals Hadamard transform coefficients definitely and, bits represents the consumption figures of mode bit number, λ represents Lagrangian constant.
6. for an intra prediction mode decision making device for HEVC efficient video coding, described intra prediction mode decision making device is to carry out predictive mode decision-making for each the prediction piece (PU) in present frame; It is characterized in that, described device is configured to:
(1) adopt 12 one-level candidate pattern at even angle interval to roughly select calculating (RMD), determine that the general orientation at optimum prediction direction place is interval;
(2) the HCost cost calculating according to RMD is arranged by ascending order candidate pattern; In two patterns of inspection HCost Least-cost, whether there is Planar pattern; If there is Planar pattern, proceed to step 3; Otherwise proceed to step 4;
(3) while there is Planar pattern in these two better predictive modes, increase by 3 secondary candidates and carry out RMD and roughly select calculating, 2 of the left and right predictive mode of DC pattern, better pre-measuring angle, sorts to candidate pattern by HCost cost size;
(4), while there is not Planar pattern in these two better predictive modes, increase by 4 secondary candidates and carry out RMD and roughly select calculating, i.e. 2 of the left and right predictive mode of 2 better pre-measuring angles,, by HCost cost size, candidate pattern is sorted;
(5) determine for the first time RMD predictive mode the selection result, obtain roughly selecting mode subset; PU size for 16 * 16,32 * 32 and 64 * 64, N=3; PU size for 4 * 4 and 8 * 8, N=2+ secondary candidate pattern number;
(6) investigate the cost size of roughly selecting each predictive mode in mode subset, if its cost HCost
mode> α * HCost
0, excluded and roughly selected mode subset; According to coding experience α=1.2;
(7) use most possible pattern (MPM) method, wherein, described in two most possible patterns are added, roughly select mode subset;
(8), to each pattern in described set, adopt rate-distortion optimization (RDO) method to obtain described optimal frames inner estimation mode.
7. intra prediction mode decision-making device as claimed in claim 6, wherein, the HCost cost that described RMD roughly selects calculating based on below:
HCost＝HSATD
mode+λ·bits
mode
Wherein, mode is one of 35 kinds of intra prediction modes of HEVC regulation, HSATD represent residual signals Hadamard transform coefficients definitely and, bits represents the consumption figures of mode bit number, λ represents Lagrangian constant.
8. one kind for executing claims the video encoder of the method described in 1～7; It is characterized in that: based on the adaptive intra prediction mode high-speed decision of classification method, there is good coding efficiency; With respect to HEVC reference software HM12.0, the inventive method code check on average increases by 0.164%, and brightness PSNR on average reduces 0.032dB, and scramble time average energy shortens 36.521%; The inventive method demonstrates good adaptability, for HEVC index plane provides technical foundation to the practical application of low complex degree.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410159107.0A CN104052994A (en) | 2014-04-14 | 2014-04-14 | Method for HEVC intra-frame prediction mode quick decision based on grading self-adaption |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410159107.0A CN104052994A (en) | 2014-04-14 | 2014-04-14 | Method for HEVC intra-frame prediction mode quick decision based on grading self-adaption |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN104052994A true CN104052994A (en) | 2014-09-17 |
Family
ID=51505311
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201410159107.0A Pending CN104052994A (en) | 2014-04-14 | 2014-04-14 | Method for HEVC intra-frame prediction mode quick decision based on grading self-adaption |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN104052994A (en) |
Cited By (18)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104301724A (en) * | 2014-10-17 | 2015-01-21 | 华为技术有限公司 | Video processing method, coding device and decoding device |
| CN104639939A (en) * | 2015-02-04 | 2015-05-20 | 四川虹电数字家庭产业技术研究院有限公司 | Optimization method for intra-frame prediction MPM (Most Probable Mode) mechanism |
| CN104639940A (en) * | 2015-03-06 | 2015-05-20 | 宁波大学 | Quick HEVC (High Efficiency Video Coding) inter-frame prediction mode selection method |
| CN104883565A (en) * | 2014-12-31 | 2015-09-02 | 乐视网信息技术（北京）股份有限公司 | Decision-making method and device for intra-frame prediction mode of high efficiency video coding |
| CN105657420A (en) * | 2015-12-30 | 2016-06-08 | 邦彦技术股份有限公司 | HEVC-oriented fast intra-frame prediction mode decision method and device |
| CN105812825A (en) * | 2016-05-10 | 2016-07-27 | 中山大学 | Image coding method based on grouping |
| CN105847794A (en) * | 2015-11-18 | 2016-08-10 | 西安邮电大学 | HEVC intra-frame prediction mode rapid selection method |
| CN106131547A (en) * | 2016-07-12 | 2016-11-16 | 北京大学深圳研究生院 | The high-speed decision method of intra prediction mode in Video coding |
| CN107371023A (en) * | 2017-08-07 | 2017-11-21 | 福建星网智慧科技股份有限公司 | The fast selecting method of luma prediction modes in a kind of frame |
| CN109862355A (en) * | 2019-03-21 | 2019-06-07 | 南京威翔科技有限公司 | Budgeting modes selection method in a kind of multi-level frame |
| CN110198442A (en) * | 2018-12-27 | 2019-09-03 | 腾讯科技（深圳）有限公司 | A kind of intra-frame prediction method, device and storage medium for Video coding |
| CN110213586A (en) * | 2019-06-10 | 2019-09-06 | 杭州电子科技大学 | A kind of VVC intra prediction angle mode fast selecting method |
| CN111918058A (en) * | 2020-07-02 | 2020-11-10 | 北京大学深圳研究生院 | Hardware-friendly intra prediction mode fast determination method, device and storage medium |
| WO2020253728A1 (en) * | 2019-06-20 | 2020-12-24 | 杭州海康威视数字技术股份有限公司 | Encoding method and device |
| CN112752105A (en) * | 2021-01-04 | 2021-05-04 | 暨南大学 | Intra-frame prediction coding complexity reduction method suitable for HEVC standard |
| CN113271461A (en) * | 2021-05-18 | 2021-08-17 | 上海大学 | Intra-frame prediction mode decision method and system based on self-adaptive cost score threshold |
| CN114205614A (en) * | 2021-12-16 | 2022-03-18 | 福州大学 | Intra-frame prediction mode parallel hardware method based on HEVC standard |
| WO2024067036A1 (en) * | 2022-09-26 | 2024-04-04 | 广州市百果园信息技术有限公司 | Decision method and apparatus for intra-frame prediction mode, and video encoding device |
-
2014
- 2014-04-14 CN CN201410159107.0A patent/CN104052994A/en active Pending
Cited By (32)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2016058357A1 (en) * | 2014-10-17 | 2016-04-21 | 华为技术有限公司 | Video processing method, encoding device and decoding device |
| CN104301724A (en) * | 2014-10-17 | 2015-01-21 | 华为技术有限公司 | Video processing method, coding device and decoding device |
| US10645382B2 (en) | 2014-10-17 | 2020-05-05 | Huawei Technologies Co., Ltd. | Video processing method, encoding device, and decoding device |
| CN104883565A (en) * | 2014-12-31 | 2015-09-02 | 乐视网信息技术（北京）股份有限公司 | Decision-making method and device for intra-frame prediction mode of high efficiency video coding |
| CN104639939B (en) * | 2015-02-04 | 2018-02-06 | 四川虹电数字家庭产业技术研究院有限公司 | A kind of optimization method of infra-frame prediction MPM mechanism |
| CN104639939A (en) * | 2015-02-04 | 2015-05-20 | 四川虹电数字家庭产业技术研究院有限公司 | Optimization method for intra-frame prediction MPM (Most Probable Mode) mechanism |
| CN104639940B (en) * | 2015-03-06 | 2017-10-10 | 宁波大学 | A kind of quick HEVC method for choosing frame inner forecast mode |
| CN104639940A (en) * | 2015-03-06 | 2015-05-20 | 宁波大学 | Quick HEVC (High Efficiency Video Coding) inter-frame prediction mode selection method |
| CN105847794A (en) * | 2015-11-18 | 2016-08-10 | 西安邮电大学 | HEVC intra-frame prediction mode rapid selection method |
| CN105657420B (en) * | 2015-12-30 | 2018-03-20 | 邦彦技术股份有限公司 | HEVC-oriented fast intra-frame prediction mode decision method and device |
| CN105657420A (en) * | 2015-12-30 | 2016-06-08 | 邦彦技术股份有限公司 | HEVC-oriented fast intra-frame prediction mode decision method and device |
| CN105812825A (en) * | 2016-05-10 | 2016-07-27 | 中山大学 | Image coding method based on grouping |
| CN105812825B (en) * | 2016-05-10 | 2019-02-26 | 中山大学 | A kind of packet-based image encoding method |
| WO2018010492A1 (en) * | 2016-07-12 | 2018-01-18 | 北京大学深圳研究生院 | Rapid decision making method for intra-frame prediction mode in video coding |
| CN106131547B (en) * | 2016-07-12 | 2018-07-03 | 北京大学深圳研究生院 | The high-speed decision method of intra prediction mode in Video coding |
| CN106131547A (en) * | 2016-07-12 | 2016-11-16 | 北京大学深圳研究生院 | The high-speed decision method of intra prediction mode in Video coding |
| CN107371023A (en) * | 2017-08-07 | 2017-11-21 | 福建星网智慧科技股份有限公司 | The fast selecting method of luma prediction modes in a kind of frame |
| CN110198442A (en) * | 2018-12-27 | 2019-09-03 | 腾讯科技（深圳）有限公司 | A kind of intra-frame prediction method, device and storage medium for Video coding |
| CN110198442B (en) * | 2018-12-27 | 2023-02-28 | 腾讯科技（深圳）有限公司 | Intra-frame prediction method, device and storage medium for video coding |
| CN109862355A (en) * | 2019-03-21 | 2019-06-07 | 南京威翔科技有限公司 | Budgeting modes selection method in a kind of multi-level frame |
| CN109862355B (en) * | 2019-03-21 | 2022-07-15 | 南京威翔科技有限公司 | Multi-level intra-frame budget mode selection method |
| CN110213586B (en) * | 2019-06-10 | 2021-03-05 | 杭州电子科技大学 | VVC intra-frame prediction angle mode rapid selection method |
| CN110213586A (en) * | 2019-06-10 | 2019-09-06 | 杭州电子科技大学 | A kind of VVC intra prediction angle mode fast selecting method |
| WO2020253728A1 (en) * | 2019-06-20 | 2020-12-24 | 杭州海康威视数字技术股份有限公司 | Encoding method and device |
| CN111918058A (en) * | 2020-07-02 | 2020-11-10 | 北京大学深圳研究生院 | Hardware-friendly intra prediction mode fast determination method, device and storage medium |
| CN111918058B (en) * | 2020-07-02 | 2022-10-28 | 北京大学深圳研究生院 | Hardware-friendly intra prediction mode fast determination method, device and storage medium |
| CN112752105A (en) * | 2021-01-04 | 2021-05-04 | 暨南大学 | Intra-frame prediction coding complexity reduction method suitable for HEVC standard |
| CN113271461A (en) * | 2021-05-18 | 2021-08-17 | 上海大学 | Intra-frame prediction mode decision method and system based on self-adaptive cost score threshold |
| CN113271461B (en) * | 2021-05-18 | 2022-09-27 | 上海大学 | Intra-frame prediction mode decision method and system based on self-adaptive cost score threshold |
| CN114205614A (en) * | 2021-12-16 | 2022-03-18 | 福州大学 | Intra-frame prediction mode parallel hardware method based on HEVC standard |
| CN114205614B (en) * | 2021-12-16 | 2023-08-04 | 福州大学 | HEVC standard-based intra-frame prediction mode parallel hardware method |
| WO2024067036A1 (en) * | 2022-09-26 | 2024-04-04 | 广州市百果园信息技术有限公司 | Decision method and apparatus for intra-frame prediction mode, and video encoding device |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN104052994A (en) | Method for HEVC intra-frame prediction mode quick decision based on grading self-adaption | |
| US11616960B2 (en) | Machine learning video processing systems and methods | |
| CN103997646B (en) | Fast intra-mode prediction mode selecting method in a kind of HD video coding | |
| CN102665079B (en) | Adaptive fast intra prediction mode decision for high efficiency video coding (HEVC) | |
| KR102304446B1 (en) | Method of encoding intra mode by choosing most probable mode with high hit rate and apparatus for the same, and method of decoding and apparatus for the same | |
| KR102070719B1 (en) | Method for inter prediction and apparatus thereof | |
| US20170094311A1 (en) | Memory-to-memory low resolution motion estimation systems and methods | |
| US9294765B2 (en) | Video encoder with intra-prediction pre-processing and methods for use therewith | |
| US11336915B2 (en) | Global motion vector video encoding systems and methods | |
| Zhang et al. | Low complexity HEVC INTRA coding for high-quality mobile video communication | |
| US10218983B2 (en) | Adapting mode decisions in video encoder | |
| US10477233B2 (en) | Predictor candidates for motion estimation search systems and methods | |
| KR20130112374A (en) | Video coding method for fast intra prediction and apparatus thereof | |
| US10225572B2 (en) | Configurable motion estimation search systems and methods | |
| Ramezanpour Fini et al. | Two stage fast mode decision algorithm for intra prediction in HEVC | |
| US11006143B2 (en) | Motion vector candidate pruning systems and methods | |
| Ramezanpour et al. | Fast HEVC I-frame coding based on strength of dominant direction of CUs | |
| Fang et al. | Fast intra mode decision for HEVC based on direction energy distribution | |
| CN104995917B (en) | A kind of adaptive movement estimation method and module | |
| CN114007078B (en) | Method, device and equipment for constructing motion information candidate list | |
| EP2899975A1 (en) | Video encoder with intra-prediction pre-processing and methods for use therewith | |
| KR20160045243A (en) | Video coding Apparatus of Planar intra prediction and method thereof | |
| Wang et al. | Revisiting local texture correlation for Rate-Distortion optimized intra coding | |
| JP6409400B2 (en) | Video encoding apparatus, method and program | |
| KR20120138430A (en) | Apparatus and method of encoding and decoding for hierarchicalintra prediction |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20140917 |
|
| WD01 | Invention patent application deemed withdrawn after publication |