CN102984521B - High-efficiency video coding inter-frame mode judging method based on temporal relativity - Google Patents

Abstract

The invention discloses a high-efficiency video coding (HEVC) inter-frame mode judging method based on temporal relativity, which comprises the steps of forecast method configuration and forecast mode selection. The forecast mode selection adopts the temporal relativity between two adjacent frames; the similarity of PU (Physical Unit) modes of a large size CU (Control Unit) of a corresponding piece in the previous frame and a small size CU in the current piece is analyzed according to the relativity; and finally, a PU mode selection method of the current CUs in various dimension is designed for the corresponding pieces in various dimensions according to the similarity. Compared with the HEVC standard in the prior art, with the adoption of the HEVC inter-frame mode judging method, the complexity of codding computation is reduced to a greater extent on the premise that the bit rate and video quality are almost unchanged.

Description

Based on the high-performance video coding inter-frame mode decision method of relativity of time domain

Technical field

The present invention relates to the Video coding inter-frame mode technical problem in field of picture communication, especially relate to a kind of high-performance video coding inter-frame mode decision method.

Background technology

Current International video coding standard is advanced video coding (H.264/AVC), this video encoding standard compare before video encoding standard, be greatly improved in video coding performance.Along with the extensive use of high-definition video technology, maximum 16 × 16 sized blocks H.264/AVC are not too applicable to encoding of HD video.For this reason, International Organization for standardization-International Electrotechnical Commission/Motion Picture Experts Group (ISO-IEC/ MPEG) and International Telecommunication Union's telecommunication standardsization tissue/Video Coding Experts group (ITU-T/VCEG) Liang great International Organization for standardization have set up Video coding and have developed jointly group (JCT-VC), and formulating international video standard of new generation, i.e. high-performance video coding (HEVC) standard.The target of HEVC standard is on the basis keeping H.264/AVC standard video quality, and make bit rate reduce half, namely compression ratio doubles.Under the effort of at home and abroad numerous scholar, more H.264/AVC the coding efficiency of HEVC video encoding standard has had higher lifting.

H.264/AVC standard adopts 16 × 16,16 × 8,8 × 16,8 × 8,8 × 4,4 × 8 and 4 × 4 equidimension blocks to carry out inter prediction, HEVC standard then adopts the coding unit (CU) of 8 × 8,16 × 16,32 × 32 and 64 × 64 sizes first to carry out piecemeal, then adopt maximum 8 kinds of different predicting unit (PU) patterns to carry out inter prediction to each CU, 8 kinds of different PU patterns are respectively PART_2N × 2N, PART_N × 2N, PART_2N × N, PART_N × N, PART_nL × 2N, PART_nR × 2N, PART_2N × nD and PART_2N × nU.PU predictive mode for the CU under often kind of size and inside thereof all carries out the calculating of rate-distortion cost, and final a kind of CU macroblock mode of the best and the PU pattern of inside thereof selected carries out Video coding.For the CU of 64 × 64 sizes, only have 3 kinds of PU patterns to select, for the CU of 32 × 32 and 16 × 16 sizes, have 7 kinds of patterns to select, for the CU of 8 × 8 sizes, then have 4 kinds of PU patterns to select.For often kind of PU pattern, encoder all needs the calculating it being carried out to rate-distortion cost, and therefore, HEVC standard brings huge Video coding computation complexity.For H.264/AVC, many scholars have been had to propose decision method between some fast frames, as on the basis of article " Fast mode decision algorithm for H.264 using statistics of rate-distortion cost " the Distribution Statistics situation of the corresponding macro block rate-distortion cost of a frame before analysis that is published in magazine " IET Electronics Letters ", propose a kind of H.264/AVC fast inter mode decision method based on adaptive threshold.The article " Fast mode decision based on mode adaptation " being published in magazine " IEEE Transactionson Circuits and Systems for Video Technology " proposes a kind of based on mode adaptive fast inter mode decision method.This algorithm is empty according to adjacent block, the video encoding behavior of time domain, constructs one based on the list of priority candidate pattern, effectively selects best inter-frame forecast mode according to this list.The article " Direct inter-mode selection for H.264 video coding using phase correlation " being published in magazine " IEEE transactions on imageprocessing ", by utilizing the phase correlation between current block and reference block to catch its motion vector, selects a kind of inter-frame forecast mode of the best from candidate pattern list according to this motion vector information.

In order to reduce the Video coding computation complexity of HEVC, scholars propose certain methods, as the article " Encoding complexity reduction by removal of N × N partition type " in HEVC motion " JCTVC-D087 " eliminates the PU pattern of PART_N × N in the CU from 16 × 16,32 × 32 and 64 × 64 sizes, only retain in the CU of 8 × 8 sizes, thus reduce Video coding computation complexity.Article " Early termination of CU encoding to reduce HEVC complexity " in motion " JCTVC-F045 " proposes as parameter cbf=0, when AC coefficient after i.e. discrete cosine transform (DCT) is complete zero, skip other PU pattern except PART_2N × 2N, thus reduce the computation complexity of Video coding.The reference software HM7.0 that above two kinds of methods have joined HEVC up-to-date has at present suffered.In addition, in current HEVC standard, for current CU, the pattern of the rate-distortion cost first detected is PART_2N × 2N, PART_2N × N and PART_N × 2N, if the rate of pattern PART_2N × N-distortion cost value is less than the rate-distortion cost value of PART_N × 2N pattern, then continue the rate-distortion cost of detecting pattern PART_2N × nU and PART_2N × nD, and the calculating of the rate-distortion cost of skip mode PART_nL × 2N and PART_nR × 2N; On the contrary, when the rate-distortion cost value of pattern PART_2N × N is greater than the rate-distortion cost of PART_N × 2N pattern, then continue the rate-distortion cost of detecting pattern PART_nL × 2N and PART_nR × 2N, and the calculating of the rate-distortion cost of skip mode PART_2N × nU and PART_2N × nD, the method also reduces the computation complexity of HEVC Video coding significantly.These methods reduce the computation complexity of HEVC Video coding all to a certain extent above, but the inter-frame forecast mode of current HEVC Video coding also exists larger redundancy in time domain.

H.264/AVC HEVC compares, video compression than in improve larger, but but increase a lot in the computation complexity of Video coding, although many scholars propose some for Fast video coding method H.264/AVC, and for the Fast video coding method of HEVC, but Fast video coding method is H.264/AVC not suitable for HEVC, and the Fast video coding method of current HEVC also exists some distances for the target being applied to real time communication.

Summary of the invention

For present situation and the deficiency of the high-performance video coding inter-frame mode decision method of prior art, of the present inventionly aim to provide a kind of high-performance video coding inter-frame mode decision method based on relativity of time domain newly, to reduce the computation complexity of HEVC Video coding, realize the target being applied to real time communication.

Basic thought of the present invention is the similitude of the PU model selection utilized between consecutive frame, according to the PU pattern that large scale CU in former frame takes, effectively select the PU pattern of current C U, skip some unlikely CU block size and PU predictive modes, reduce the PU pattern of required traversal, thus realize the quantity of slip-distortion cost calculating, finally reach the object of the computation complexity reducing HEVC Video coding.

High-performance video coding inter-frame mode decision method based on relativity of time domain provided by the invention, comprise prediction mode configuration and predictive mode selection, in prediction mode configuration, CU is split the degree of depth and is not more than 4, PU adopts symmetry and asymmetric integrated forecasting pattern or only adopts symmetrical predictive mode, in predictive mode is selected, rate total for current depth CU-distortion cost sum and the total rate-distortion cost sum of last layer CU are compared, if less than upper strata, quad-tree partition is then taked to become the CU of 4 more next layer depth further, otherwise termination quad-tree partition, described predictive mode is selected to comprise the following steps:

(1) size of current C U piecemeal former frame correspondence position CU piecemeal is detected, if current C U block size is less than the size of corresponding CU piecemeal, then enter step (2) below, otherwise all PU patterns of traversal current C U piecemeal, and take quad-tree partition to become 4 darker one deck CU piecemeals, said process is repeated to more further each CU piecemeal;

(2) whether the PU pattern judging the corresponding CU piecemeal of former frame is PART_2N × 2N, and if so, then current C U piecemeal only detects the rate-distortion cost of the PU pattern of PART_2N × 2N, and enters step (6) below, otherwise enters step (3) below;

(3) whether the PU pattern judging the corresponding CU piecemeal of former frame is PART_nL × 2N or PART_nR × 2N, if, then current C U piecemeal only detects the rate-distortion cost of PART_N × 2N and PART_2N × 2N two kinds of PU patterns, and enter step (6) below, otherwise enter step (4) below;

(4) whether the PU pattern judging the corresponding CU piecemeal of former frame is PART_2N × nU or PART_2N × nD, if, then current C U piecemeal only detects the rate-distortion cost of PART_2N × N and PART_2N × 2N two kinds of PU patterns, and enter step (6) below, otherwise enter step (5) below;

(5) detect rate-distortion cost that current C U piecemeal detects all PU patterns, and enter step (6) below;

(6) judge that whether the size of current C U piecemeal is 1/4 of former frame corresponding CU block size, if so, then current C U piecemeal no longer carries out quad-tree partition; Otherwise taked by current C U piecemeal quad-tree partition to become 4 darker one deck CU piecemeals further, step (1) process is repeated to more further each CU piecemeal.

In technique scheme, CU is split the degree of depth and is preferably 2 ~ 4, more preferably 4.In the adoptable symmetry of PU and asymmetric integrated forecasting pattern with only adopt in symmetrical predictive mode, preferentially adopt symmetrical and asymmetric integrated forecasting pattern.

In technique scheme, described rate-distortion cost is determined by following formula:

J _mode=(SAD _luma+w _chroma×SAD _chroma)+λ _mode×B _mode

J in formula _modefor rate-distortion cost, SAD _lumafor the mean square deviation of original image brightness and predicted picture brightness, SAD _chromafor the mean square deviation of original image colourity and predicted picture colourity, w _chromafor the weights of chromatic distortion, λ _moderepresent Lagrange multiplier, B _moderepresent number of coded bits in this mode.

The video encoder performing the above-mentioned high-performance video coding inter-frame mode decision method based on relativity of time domain can be worked out according to said method of the present invention.

The present invention completes based on following thinking analysis:

When the PU pattern of corresponding blocks CU is PART_2N × 2N, illustrate that the texture of corresponding blocks should be smoother or in region residing for it, all Moving Objects have identical motion vector, according to the similitude of height between adjacent two frames, current block region also should have similar attribute, therefore, for current C U piecemeal, only need detect PU pattern PART_2N × 2N.

When the PU pattern of corresponding blocks CU is PART_2N × N or PART_N × 2N, in this case, according to the similitude between adjacent two frames, the separator bar of PU is not had in the corresponding blocks that CU now in current block is corresponding, so, now can not infer that corresponding blocks region is that level and smooth or wherein Moving Objects all has identical motion vector.Therefore, for current C U piecemeal, not having can the PU pattern of reference, must travel through all PU patterns.

When the PU pattern of corresponding blocks CU is PART_nL × 2N or PART_nR × 2N, according to the relativity of time domain between adjacent two frames and the direction of motion in time domain, illustrate that on the one hand the Moving Objects in now current block has larger probability to be divided into left and right two parts, on the other hand, now there is the separator bar of the PU pattern PART_N × 2N of part CU in current block by corresponding for the separator bar of PU pattern PART_nL × 2N or the PART_nR × 2N with corresponding blocks CU, moreover, according to the correlation of PU pattern in current block (with the current block in the present frame of corresponding CU block same position and size) between each layer and the PU pattern dependency with adjacent C U under CU layer, in current block each layer CU PU pattern in, occur that PART_N × 2N probability is higher, therefore, current block should detect PU pattern PART_N × 2N.In addition, in order to the continuity between each layer CU and guarantee video coding performance, PU pattern PART_2N × 2N also should take into account, and therefore, in this case, altogether only travels through 2 kinds of PU pattern: PART_N × 2N and PART_2N × 2N.In like manner, when the PU pattern of corresponding blocks CU is PART_2N × nU or PART_2N × nD, now illustrate that the Moving Objects in current block has larger probability to be divided into left and right two parts, therefore, it is PART_2N × N and PART_2N × 2N that current block should detect PU pattern.

Finally, when corresponding blocks CU block size is 2N × 2N, and when in current block, minimum CU is of a size of N/4 × N/4, according to the correlation of time domain between adjacent two frames, no matter which kind of the PU pattern of corresponding CU piecemeal is, the cut-off rule being of a size of the PU pattern of the CU of N/4 × N/4 in current block all can not be corresponding with the cut-off rule of the PU pattern of corresponding CU, namely now current C U block size unlikely occurs N/4 × N/4, therefore, for the CU of this size, its rate-distortion cost can not be detected to save Video coding computation complexity.

Compare HEVC video encoding standard, method of the present invention can on the basis reducing Video coding computation complexity by a relatively large margin, Video coding compression ratio and video quality loss very little.The fundamental basis of video-frequency compression method is the amount of information reached with the original whole video of less data representation by reducing various correlation in video, method of the present invention is analyzing the correlation in the time domain in video between adjacent two frames between PU pattern, by judging the PU pattern of the corresponding CU of former frame, and the correlation in current block between adjacent C U piecemeal, the best PU pattern of judgement current block, thus skipped other PU pattern, from the correlation angle of PU pattern between consecutive frame, the inventive method eliminates the redundancy between PU pattern, and reasonably redundancy is removed, the Video coding calculation task carried out needed for this redundancy section can not only be removed, also can not lose the amount of information of video simultaneously, therefore the decline of video compression ratio and video quality can not also substantially be caused.

What the inventive method was improved is the place of whole Video coding computational process complexity most critical.In whole video coding process, according to different configuration links, there is some difference more than 50%(for the ratio shared by computation complexity of estimation (comprising integer pixel estimation and fractional-pel motion estimation)), the inventive method is it is crucial that according to the relativity of time domain between adjacent two frames, skip the testing process of the PU pattern of the unlikely appearance in current block in each layer CU block, namely each PU piecemeal under this PU pattern is carried out to the calculating detection of rate-distortion cost, thus select a minimum PU pattern of wherein rate-distortion cost, and in the calculating of rate-distortion cost, estimation is coding computational process wherein the most consuming time, skip several PU pattern and namely mean the amount of calculation skipping several estimation, therefore, in computation complexity, the point that the inventive method is set about is most critical improvements in Video coding computational process.

The inventive method keeping reducing on the basis of computation complexity, additionally can not increase hardware implementation cost.Finally all to embed hardware device in a lot of situation of video coding technique, comprise FPGA and DSP etc., therefore, require all higher for the code computing cost of improving one's methods and required data storage hardware cost.The inventive method needs the code of increase little, mainly comprise several judgement statement, in the memory needed for hardware, due to the PU pattern that the object judged in the inventive method is correspondence position CU piecemeal in former frame, and the information of these patterns just stored in a stream originally, and the inventive method does not bring extra call data storage, therefore, if the inventive method is applied to hardware device, extra cost can not be increased to the manufacture of hardware device, can also power consumption be saved simultaneously.

Accompanying drawing explanation

Fig. 1 compares schematic diagram based on the HEVC fast inter mode decision method of time domain specification and the CU method of partition of HM7.0 video encoding standard, wherein (a) CU method of partition that is HM7.0 video encoding standard, (b) is CU method of partition in the HEVC fast inter mode decision method based on relativity of time domain;

Fig. 2 compares schematic diagram based on the HEVC fast inter mode decision method of time domain specification and the PU Forecasting Methodology of HM7.0 video encoding standard, wherein (a) PU Forecasting Methodology that is HM7.0 video encoding standard, (b) is PU Forecasting Methodology in the HEVC fast inter mode decision method based on relativity of time domain;

Fig. 3 is the flow chart of the HEVC fast inter mode decision method based on relativity of time domain.

Embodiment

Below in conjunction with embodiment, the present invention is described in further detail; it is important to point out; following embodiment is only for the present invention is described further; limiting the scope of the invention can not be interpreted as; affiliated art skilled staff is according to foregoing invention content; some nonessential improvement are made to the present invention and adjustment is specifically implemented, still should belong to protection scope of the present invention.

1. open the program of two algorithms simultaneously and set identical configuration file, reference software selects HM7.0, and quantization step (QP) value gets 27 and 32 respectively.The method of the reference software algorithm HM7.0 with HEVC video encoding standard compares by the present invention.And to its three kinds of video coding performances: (wherein PSNR embodies the objective video quality of video for bit rate, Y-PSNR (PSNR) and video encoding time, video encoding time embodies the computation complexity of coding), compare analysis, the gap comparing performance is evaluated by following three indexs:

$\mathrm{\ΔBitrate}=\frac{{\mathrm{Bitrate}}_{\mathrm{pro}}-{\mathrm{Bitrate}}_{\mathrm{ref}}}{{\mathrm{Bitrate}}_{\mathrm{ref}}}\×100\%$

$\mathrm{\ΔPSNR}={\mathrm{PSNR}}_{\mathrm{pro}}-{\mathrm{PSNR}}_{\mathrm{ref}}$

$\mathrm{\ΔTime}=\frac{{\mathrm{Time}}_{\mathrm{pro}}-{\mathrm{Time}}_{\mathrm{ref}}}{{\mathrm{Time}}_{\mathrm{ref}}}\×100\%$

Wherein Bitrate _pro, PSNR _proand Time _probe respectively the bit rate of algorithm of the present invention, PSNR and video encoding time, Bitrate _ref, PSNR _refand Time _refbe respectively the bit rate of HM7.0 canonical algorithm, PSNR and video encoding time, Bitrate, PSNR and Time are respectively the difference of bit rate, PSNR and video encoding time between algorithm of the present invention and HM7.0 canonical algorithm.

2. in HEVC video coding technique, PU predictive mode can take symmetrical and asymmetrical integrated forecasting pattern, also symmetrical predictive mode can only be taked, invent all effective in both cases, but take symmetry and asymmetrical integrated forecasting pattern can reduce more Video coding computation complexity, can obtain better algorithm effect, therefore the present invention takes integrated forecasting pattern.

3. the object of coding is the HEVC test video of standard, their title, resolution and frame per second are respectively: Fourpeople(1280 × 720,60 frames/second), Johnny(1280 × 720,60 frames/second), KristenandSara(1280 × 720,60 frames/second), Cactus(1920 × 1080,50 frames/second), Kimono1(1920 × 1080,24 frames/second) and ParkScene(1920 × 1080,24 frames/second).

4. the video sequence that input 2 is identical;

5. respectively Video coding is carried out to 2 identical video sequences;

6. utilize HEVC video encoder HM7.0 under HEVC mode, to carry out Video coding to video sequence;

7. algorithm of the present invention is selected according to the PU pattern of the PU pattern of former frame corresponding blocks to current C U piecemeal;

8. in predictive mode is selected, rate total for current depth CU-distortion cost sum and the total rate-distortion cost sum of last layer CU are compared, if less than upper strata, quad-tree partition is then taked to become the CU of 4 more next layer depth further, otherwise termination quad-tree partition, concrete predictive mode is selected as follows:

(1) size of current C U piecemeal former frame correspondence position CU piecemeal is detected, if current C U block size is less than the size of corresponding CU piecemeal, then enter step (2) below, otherwise all PU patterns of traversal current C U piecemeal, and take quad-tree partition to become 4 darker one deck CU piecemeals, said process is repeated to more further each CU piecemeal;

(2) whether the PU pattern judging the corresponding CU piecemeal of former frame is PART_2N × 2N, and if so, then current C U piecemeal only detects the rate-distortion cost of the PU pattern of PART_2N × 2N, and enters step (6) below, otherwise enters step (3) below;

(3) whether the PU pattern judging the corresponding CU piecemeal of former frame is PART_nL × 2N or PART_nR × 2N, if, then current C U piecemeal only detects the rate-distortion cost of PART_N × 2N and PART_2N × 2N two kinds of PU patterns, and enter step (6) below, otherwise enter step (4) below;

(4) whether the PU pattern judging the corresponding CU piecemeal of former frame is PART_2N × nU or PART_2N × nD, if, then current C U piecemeal only detects the rate-distortion cost of PART_2N × N and PART_2N × 2N two kinds of PU patterns, and enter step (6) below, otherwise enter step (5) below;

(5) detect rate-distortion cost that current C U piecemeal detects all PU patterns, and enter step (6) below;

(6) judge that whether the size of current C U piecemeal is 1/4 of former frame corresponding CU block size, if so, then current C U piecemeal no longer carries out Quadtree Partition division; Otherwise current C U piecemeal quad-tree partition is become 4 CU piecemeals, repetition step (1) is carried out to each CU piecemeal.

9., in mode selection processes, the formula of rate distortion costs is as follows:

J _mode=(SAD _luma+w _chroma×SAD _chroma)+λ _mode×B _mode

J in formula _modefor rate-distortion cost, SAD _lumafor the mean square deviation of original image brightness and predicted picture brightness, SAD _chromafor the mean square deviation of original image colourity and predicted picture colourity, w _chromafor the weights of chromatic distortion, λ _moderepresent Lagrange multiplier, B _moderepresent number of coded bits in this mode.

The distortion SAD of brightness and colourity _lumaand SAD _chromacan be drawn by following two formulas respectively:

${\mathrm{SAD}}_{\mathrm{luma}}=\underset{i,j}{\mathrm{\Σ}}\left|{\mathrm{Diff}}_{\mathrm{luma}}(i,j)\right|$

${\mathrm{SAD}}_{\mathrm{chroma}}=\underset{i,j}{\mathrm{\Σ}}\left|{\mathrm{Diff}}_{\mathrm{chroma}}(i,j)\right|$

Wherein Diff _lumaand Diff _chromabe respectively:

Diff _luma(i,j)=BlockA _luma(i,j)-BlockAB _luma(i,j)

Diff _chroma(i,j)=BlockA _chroma(i,j)-BlockAB _chroma(i,j)

Wherein BlockA _lumaand BlockB _lumabeing respectively coordinate position in encoding block and prediction block is the pixel brightness value on (i, j), BlockA _chromaand BlockB _chromabeing respectively coordinate position in encoding block and prediction block is the pixel color angle value on (i, j).

Chromatic distortion weight w _chromacan be drawn by following formula:

$w_{\mathrm{chroma}}=2^{(\mathrm{QP}-{\mathrm{QP}}_{\mathrm{chroma}})/3}$

Wherein QP and QP _chromabe respectively the QP value of brightness and colourity.

Lagrange multiplier λ _modecan be drawn by following formula:

λ _mode=2 ^(QP-12)/3

10. in algorithm of the present invention, CU split the degree of depth can get 2 ~ 4, but segmentation the degree of depth get larger, then the PU predictive mode skipped in the method will be more, reduce more Video coding computation complexity the most at last.Therefore, the CU segmentation degree of depth of the present invention gets 4.

Video sequence after 11. two program difference output video codings and respective bit rate, PSNR value and total video encoding time, the result of above-mentioned 3 indexs is as shown in table 1-3, statistics display algorithm of the present invention rises 0.13-1.05% than HEVC standard in bit rate, and it seems generally, more more than what will rise a little under little QP value under large QP value, in video quality PSNR value, reduce 0.00-0.06dB, in Video coding computation complexity, reduce 22.38-58.36%.Come as a whole, algorithm of the present invention is compared with HEVC video encoding standard, under video compression ratio (being embodied by bit rate decline degree) and the very little prerequisite of video quality loss, reduce the computation complexity (see table 1 ~ 3) of Video coding greatly.

Table 1 algorithm of the present invention compares with bit rate between HM7.0 canonical algorithm

Table 2 algorithm of the present invention compares with PSNR value between HM7.0 canonical algorithm

Table 3 algorithm of the present invention compares with video encoding time between HM7.0 canonical algorithm

Claims (4)

1. the high-performance video coding inter-frame mode decision method based on relativity of time domain, comprise prediction mode configuration and predictive mode selection, in prediction mode configuration, coding unit CU is split the degree of depth and is not more than 4, predicting unit PU adopts symmetry and asymmetric integrated forecasting pattern or only adopts symmetrical predictive mode, in predictive mode is selected, rate total for current depth coding unit CU-distortion cost sum and the total rate-distortion cost sum of last layer coding unit CU are compared, if less than upper strata, quad-tree partition is then taked to become the coding unit CU of 4 more next layer depth further, otherwise termination quad-tree partition, it is characterized in that described predictive mode is selected to comprise the following steps:

(1) size of current coded unit CU piecemeal former frame correspondence position coding unit CU piecemeal is detected, if current coded unit CU block size is less than the size of corresponding coding unit CU piecemeal, then enter step (2) below, otherwise all predicting unit PU patterns of traversal current coded unit CU piecemeal, and take quad-tree partition to become 4 darker one deck coding unit CU piecemeals, said process is repeated to more further each coding unit CU piecemeal;

(2) whether the predicting unit PU pattern judging former frame corresponding coding unit CU piecemeal is PART_2N × 2N, if, then current coded unit CU piecemeal only detects the rate-distortion cost of the predicting unit PU pattern of PART_2N × 2N, and enter step (6) below, otherwise enter step (3) below;

(3) whether the predicting unit PU pattern judging former frame corresponding coding unit CU piecemeal is PART_nL × 2N or PART_nR × 2N, if, then current coded unit CU piecemeal only detects the rate-distortion cost of PART_N × 2N and PART_2N × 2N two kinds of predicting unit PU patterns, and enter step (6) below, otherwise enter step (4) below;

(4) whether the predicting unit PU pattern judging former frame corresponding coding unit CU piecemeal is PART_2N × nU or PART_2N × nD, if, then current coded unit CU piecemeal only detects the rate-distortion cost of PART_2N × N and PART_2N × 2N two kinds of predicting unit PU patterns, and enter step (6) below, otherwise enter step (5) below;

(5) detect rate-distortion cost that current coded unit CU piecemeal detects all predicting unit PU patterns, and enter step (6) below;

(6) judge that whether the size of current coded unit CU piecemeal is 1/4 of former frame corresponding coding unit CU block size, if so, then current coded unit CU piecemeal no longer carries out quad-tree partition; Otherwise taked by current coded unit CU piecemeal quad-tree partition to become 4 darker one deck coding unit CU piecemeals further, step (1) process is repeated to more further each coding unit CU piecemeal.

2., as claimed in claim 1 based on the high-performance video coding inter-frame mode decision method of relativity of time domain, it is characterized in that coding unit CU splits the degree of depth is 2 ~ 4.

3., as claimed in claim 2 based on the high-performance video coding inter-frame mode decision method of relativity of time domain, it is characterized in that coding unit CU splits the degree of depth is 4.

4. the high-performance video coding inter-frame mode decision method based on relativity of time domain as described in one of claims 1 to 3, is characterized in that described rate-distortion cost is determined by following formula:

J _mode＝(SAD _luma+w _chroma×SAD _chroma)+λ _mode×B _mode

J in formula _modefor rate-distortion cost, SAD _lumafor the mean square deviation of original image brightness and predicted picture brightness, SAD _chromafor the mean square deviation of original image colourity and predicted picture colourity, w _chromafor the weights of chromatic distortion, λ _moderepresent Lagrange multiplier, B _moderepresent number of coded bits in this mode.