CN103188496A - Fast motion estimation video encoding method based on motion vector distribution forecast - Google Patents

Abstract

The invention belongs to the field of video compression encoding and discloses a fast motion estimation video encoding method based on motion vector distribution forecast. The fast motion estimation video encoding method comprises the following steps of: firstly extracting brightness information of a current macroblock from original video data, designing a concise motion estimation search template aiming at integer-pixel motion vector distribution characteristic search, reasonably distributing search points, then forecasting motion vector distribution, and self-adaptively selecting to carry out small-scale search in a corresponding area of the search template according to a forecasting result; and judging whether a motion vector is 0 in the motion estimation search process and taking the motion vector as a criterion of skipping template search. Compared with a motion estimation search algorithm and other improved algorithms adopted in a H.264 video encoding standard, the fast motion estimation video encoding method disclosed by the invention can be used for effectively accelerating the motion estimation search process, shortening the time spent on motion estimation, strictly controlling the increase of code rate, ensuring the better quality of a reconstructed image and achieving the fast motion estimation coding.

Description

Method for coding quick movement estimation video based on the motion vector distribution prediction

Technical field

The invention belongs to the video compression coding field, relate to a kind of quick video motion estimated coding method based on the motion vector distribution prediction.

Background technology

Unite promulgation in 2003 by International Telecommunications Union (ITU-T) and International Standards Organization (ISO/IEC) and proposed video encoding standard of new generation H.264/AVC.H.264 You Yi compression performance and video quality make it be widely used in actual life, comprise that digital television broadcasting, multi-media SMS, the transmission of Internet video Streaming Media communicate by letter etc. with real-time video.H.264, coding standard than before improves in the following aspects: adopt 1/4 pixel precision estimation; Support 16 * 16,16 * 8,8 * l6,8 * 8,8 * 4,4 * 8,4 * 4 be the prediction mode of totally 7 kinds of sized blocks; Support the multiframe reference; Adopt 4 * 4 integer transforms to replace 8 * 8 dct transform; Optional 2 kinds of entropy coded systems: based on context-adaptive variable-length encoding (context Adaptive Variable Length Coding, be called for short CAVLC) and based on context adaptive binary arithmetic coding (Context Adaptive Binary Arithmetic Coding is called for short CABAC); Adopt the rate-distortion optimization technology, take all factors into consideration coding efficiency and code check in the cataloged procedure and weigh.Though H.264 very big improvement has been arranged aspect coding efficiency, but has been that cost exchanges for to sacrifice encoder complexity.H.264 high complexity is calculated and is made and coding length consuming time had a strong impact on its real-time performance, so most important for the research that promotes coding rate aspect H.264.

Correlative study shows, H.264 the inter prediction encoding that comes from mostly consuming time of encoding, wherein estimation (Motion Estimation is called for short ME) process has accounted for 60%-80% of coding total time, is the key component that increases computational complexity and influence coding efficiency.In recent years, researchers are devoted to motion estimation process is optimized always, when guaranteeing coding efficiency, shorten the estimation time as far as possible.

H.264 adopt the block matching motion algorithm for estimating for estimation in, its principle as shown in Figure 1.At first video one frame is divided into some M * N piece, supposes that all pixels are all done identical translational motion in the piece.If the former frame field of search is (M+2W _x, N+2W _y), the displacement of present frame piece and former frame piece is that (i, j), if can find the former frame piece that mates with the present frame piece in the field of search, (i j) is needed motion vector MV (Motion vector) to this d to d.In block matching method, adopt absolute difference and SAD (Sum of Absolute Difference) to weigh current block and the degree of correlation of prediction between the piece as correlation function usually, sad value is more little, represents that then correlation is more strong, and the two is coupling more.

In existing block matching motion algorithm for estimating, that search precision is the highest is full-search algorithm (FS).Because FS algorithm computation complexity is too high, be unfavorable for real-time application, the researcher has proposed three-step approach (TSS), four step rule (FSS) and hexagon method new searching algorithms such as (HEXBS) in succession for this reason, but they still exist search point many, easily are absorbed in the deficiency of local optimum.Along with going deep into of research, the new algorithm that proposes at temporal correlation and human-eye visual characteristic has had very great development, asymmetric cross multilayer hexagon search algorithm (Unsymmetrical-Cross Multi-Hexagon Search is called for short UMHexagonS) is typically arranged.Than full-search algorithm, this algorithm can reduce for 90% estimation time, and Y-PSNR on average descends less than 0.05dB, and it is constant substantially to keep code check.

Application number be CN201010140709.3 patent disclosure a kind of method for coding quick movement estimation video based on the macroblock motion vector tagsort, be called NUMHexagonS(New-UMHexagonS) algorithm, its motion estimation search scheme is as shown in Figure 2.This algorithm has had significant improvement at the UMHexagonS algorithm, but still exists and to fail to take full advantage of the motion vector distribution feature and carry out search strategy and formulate, divide for large-scale search pattern region of search meticulous and to have ignored motion vector be 0 o'clock deficiencies such as search characteristics inadequately.

Summary of the invention

At the above-mentioned problems in the prior art, the present invention proposes a kind of new motion estimation search algorithm, keeping under low code check and the high-quality prerequisite, search point reaches the shortening video scramble time in the motion estimation process by reducing, and improves the code efficiency purpose.

The technical solution used in the present invention is: based on the method for coding quick movement estimation video of motion vector distribution prediction, the shortcoming that has the search point waste at the extensive search template, unsymmetrical cross searching template and non-homogeneous hexagonal mesh search pattern are redesigned, and before unsymmetrical cross searching and the search of non-homogeneous hexagonal mesh, carry out the motion vector distribution prediction respectively, select corresponding region of search adaptively, judge at searching period whether motion vector is 0, add the premature termination criterion, and then realization is characterized in that specifically comprising the steps: to the coding quick movement estimation video of inter macroblocks

Step 1 is extracted the monochrome information of current predicted macro block as coded object from current video frame;

Step 2 is determined twofold motion estimating searching template, carries out search point and distributes, and method is as follows:

(1) unsymmetrical cross searching template: formed by 4 regions of search, distribute 4 search points vertically respectively, distribute 8 search points about level respectively;

(2) non-homogeneous hexagonal mesh search pattern: constitute 32 regions of search jointly by 4 layers and 8 directions, the non-homogeneous distribution of counting is carried out in each zone, increase along with search radius, the expansion of hunting zone, distribute search point to increase progressively from the inside to the outside, have 62 of search points in the horizontal direction ± 45 in ° zone, have 22 of search points in the vertical direction ± 45 in ° zone;

Step 3, determine the initial search point of high precision: adjudicate foundation by Lagrangian rate-distortion optimization (RDO-Rate Distortion Optimization) function as estimation, optimum Match prediction piece and optimal motion vector on the selection rate distortion sense make the Bit Allocation in Discrete minimum of motion vector and residual coding.Utilize Lagrangian rate distortion criterion to select the optimal motion vector problem can be described as:

J _motion(mv,ref|λ _motion)＝SAD[s,r(ref,mv)]+λ _motion[R(mv-pred)+R(ref)]

Wherein, J _MotionRate distortion costs value for the motion vector of current prediction

S is the current macro block pixels value, mv is current motion vector, pred is motion vectors, ref is reference frame, r (ref, mv) be the pixel value of reference macroblock, R is that motion vector carries out the bit number that differential coding consumes, and comprises the differential coding bit number of motion vector and its predicted value and the number of coded bits of reference frame; λ _MotionBe Lagrange multiplier, SAD(Sum of Absolute Difference) be between current block and reference block pixel absolute error and:

$\mathrm{SAD}[s,r(\mathrm{ref},\mathrm{mv})]=\underset{x=1}{\overset{B_1}{\mathrm{\Σ}}}\underset{y=1}{\overset{B_2}{\mathrm{\Σ}}}|s(x,y)-r(x-m_x,u-m_y)|$

Wherein, B ₁And B ₂Horizontal pixel and the vertical pixel of representing piece respectively, according to different inter-frame forecast modes, its value can be 16,8, and 4; (x y) is the current macro block pixels value to s; (x y) is the pixel value of reference macroblock, m to r _xAnd m _yThe displacement of representing level and vertical direction respectively.

Utilize Lagrangian rate distortion criterion to select the problem of optimization model to be expressed as:

J _mode(s,c,MODE|λ _mode)＝SSD(s,c,MODE|QP)+λ _mode×R(s,c,MODE|QP) （1）

Wherein, MODE represents a kind of interframe encoding mode of current macro; J _Mode(s, c, MODE| λ _Mode) rate distortion costs value under the expression MODE pattern

S is original vision signal; C is the reconstructed video signal behind the employing MODE pattern-coding; λ _ModeBe Lagrange multiplier; (s, c MODE|QP) are the total number of bits that comprise macro block header, motion vector and all DCT block messages relevant with MODE and quantization parameter to R, and it is by obtaining behind the coding that piece is carried out reality, so its operand is bigger; QP is the coded quantization step-length; (MODE|QP) (Sum of Square Difference) is the difference of two squares sum between primary signal and the reconstruction signal to SSD, that is: for s, c

$\mathrm{SSD}(s,c,\mathrm{MODE}|\mathrm{QP})=\underset{i=1,j=1}{\overset{B1,B2}{\mathrm{\Σ}}}{[S_Y(x,y)-G_Y(x,y,\mathrm{MODE}|\mathrm{QP})]}^2$

$+\underset{i=1,j=1}{\overset{B1,B2}{\mathrm{\Σ}}}{[S_U(x,y)-G_U(x,y,\mathrm{MODE}|\mathrm{QP})]}^2+\underset{i=1,j=1}{\overset{B1,B2}{\mathrm{\Σ}}}{[S_V(x,y)-C_V(x,y,\mathrm{MODE}|\mathrm{QP})]}^2$

Wherein, B ₁And B ₂Represent horizontal pixel and the vertical pixel of piece respectively, its value can be 16,8, and 4; S _Y(x y) is the value of source macro block brightness information; C _Y(MODE|QP) value of the monochrome information of macro block is rebuild in expression for x, y; S _U, S _VAnd C _U, C _VThe value of representing corresponding colour difference information respectively.

Specifically may further comprise the steps:

(1) adopt multiple prediction mode to carry out the starting point prediction:

1) carries out the median prediction MV of spatial domain _{Pred_space}: utilize the correlation between adjacent macroblocks, the motion vector of current predicted macroblock is obtained by the motion-vector prediction of adjacent piece around known;

2) carry out the upper strata piece prediction MV that many sized blocks are divided _{Pred_uplayer}: utilize interframe movement to estimate the macroblock partitions feature, current predicted macroblock motion vector is obtained co-located, upper level in proper order and is twice the motion vector of sized blocks from the hierarchical search of 7 patterns;

3) carry out the reference frame motion-vector prediction MV of time-domain _{Pred_ref}: utilize the motion vector of current macro in different reference frames to have correlation, the current macro motion vector is predicted by a certain percentage by the motion vector of current block in the reference frame before and is obtained.

(2) the rate distortion costs value with these three kinds of motion vector prediction mode points pointed is designated as space_pred respectively _Mincost, uplayer_pred _MincostAnd ref_pred _MincostTo have the motion vector points point of minimum rate distortion costs value as initial search point, the optimal match point of this step is wanted the terminal point of motion vectors as next step.

Step 4, judge whether motion vector is 0:

The motion vector that obtains the current macro optimal match point in the previous step is designated as (MV _x, MV _y), if MV _x=0 and MV _y=0, then be considered as this moment motion vector be 0, skips steps five is directly carried out step 8.

Step 5 is carried out the motion vector direction prediction to current macro: according to next step unsymmetrical cross searching template characteristics current predicted macroblock motion vector is divided into four direction, i.e. horizontal positive direction, horizontal negative direction, vertical positive direction and vertical negative direction.Current predicted macroblock motion vector distribution direction prediction step is as follows:

(1) according to current macro motion vector position calculation directioin parameter:

Current macro motion vector coordinate (MV _x, MV _y), the motion vector direction is with its direction vector MV=(MV _x, MV _y) expression.Calculate

Value.

(2) judge current predicted macroblock motion vector direction:

When k＜1, according to MV _xState is judged: work as MV _xJudged that current macro motion vector direction was horizontal positive direction at＞0 o'clock, work as MV _xJudged that current macro motion vector direction was horizontal negative direction at＜0 o'clock;

Work as k〉1 or MV _x=0 o'clock, according to MV _yState is judged: work as MV _yJudged that current macro motion vector direction was vertical positive direction at＞0 o'clock, work as MV _yJudged that current macro motion vector direction was vertical negative direction at＜0 o'clock;

Step 6, carry out unsymmetrical cross searching: the vertical search scope is half of horizon scan scope, step-length is two pixels between the adjacency search point.Select the direction of search adaptively according to current macro motion vector direction prediction result in the 5th step: when being judged to be horizontal positive direction, 8 search points on the horizontal positive direction of search initial point; When being judged to be horizontal negative direction, 8 search points on the horizontal negative direction of search initial point; When being judged to be vertical positive direction, 4 search points on the vertical positive direction of search initial point; When being judged to be vertical negative direction, 4 search points on the vertical negative direction of search initial point.The optimum Match of determining in this step is named a person for a particular job and is wanted the terminal point of motion vectors as next step.

Step 7 is judged current macro motion vector size: current predicted macroblock motion vector size is divided into motion vector is big, motion vector is medium and less three ranks of motion vector.The determination methods of current predicted macroblock motion vector size is as follows:

(1) determines pred according to the prediction mode of initial search point _MincostAnd RD _MincostValue:

If in step 3, adopt the time prediction mode, then: pred _Mincost=ref_pred _Mincost

If in step 3, adopt the spatial prediction mode, then: pred _Mincost=space_pred _Mincost

If in step 3, adopt upper strata piece prediction mode, then: pred _Mincost=uplayer_pred _Mincost

RD _MincostFor by calculating the minimum rate distortion costs value of current predicted macroblock under the MODE pattern in the formula (1).

(2) calculate the motion vectors parameter:

The motion vectors lower threshold is:

(1+γ)pred _mincost

The motion vectors upper limit threshold is:

(1+δ)pred _mincost

The motion vectors factor is:

$\mathrm{\γ}=\frac{\mathrm{Bsize}\left[\mathrm{blocktype}\right]}{{{\mathrm{pred}}_{\min \cos t}}^2}-{\mathrm{\α}}_1\left[\mathrm{blocktype}\right]$

$\mathrm{\δ}=\frac{\mathrm{Bsize}\left[\mathrm{blocktype}\right]}{{{\mathrm{pred}}_{\min \cos t}}^2}-{\mathrm{\α}}_2\left[\mathrm{blocktype}\right]$

Wherein, Bsize[blocktype] size of the current predicted macroblock of expression, its value can be 16,8,4,

α ₁[blocktype]＝[-0.23,-0.23,-0.23,-0.25,-0.27,-0.27,-0.28]

α ₂[blocktype]＝[-2.39,-2.40,-2.40,-2.41,-2.45,-2.45,-2.48]

(3) judge current predicted macroblock motion vector size:

Work as RD _Mincost≤ (1+ γ) pred _Mincost, when namely minimum rate distortion costs value is less than the motion vectors lower threshold, judge that the current macro motion vector is less, namely movement degree is mild;

As (1+ γ) pred _Mincost＜RD _Mincost＜(1+ δ) pred _Mincost, when namely minimum rate distortion costs value is between motion vectors lower threshold and motion vector upper limit threshold, judge that the current macro motion vector is medium, namely movement degree is medium;

Work as RD _Mincost〉=(1+ δ) pred _Mincost, when namely minimum rate distortion costs value is greater than the motion vectors upper limit threshold, judge that the current macro motion vector is bigger, namely movement degree is more violent.

Step 8 is optionally carried out 5 * 5 pixels and is searched for entirely: optionally carries out 5 * 5 pixels according to prediction current macro motion vector size result in the step 7 and search for entirely.Only motion vector hour or motion vector be 0 o'clock just centered by current future position, in 4 * 4 zones around it, carry out the full search of 5 * 5 search points; Otherwise do not carry out 5 * 5 search entirely, directly enter step 9.

Step 9, current macro is judged again whether motion vector is 0, and carry out the prediction of motion vector direction prediction and motion vector size: the optimal match point in the step 8 is predicted again, judged that wherein whether motion vector is 0 identical with above-mentioned steps four and step 7 with motion vector size Forecasting Methodology.If motion vector is 0, then skips steps ten, directly carry out step 11; If motion vector is not 0, then obtains the current macro motion vector distribution by motion vector direction prediction and the prediction of motion vector size and predict the outcome.According to next step non-homogeneous multilayer hexagonal mesh search pattern characteristics current predicted macroblock motion vector is divided into eight directions:

(-22.5°,22.5°]，(157.5°,202.5°]，(22.5°,67.5°]，(-157.5°,-112.5°]，

(67.5°,112.5°],(-112.5°,-67.5°],(112.5°,157.5°],(-67.5°,-22.5°]；

The method of motion vector direction prediction is as follows:

(1) according to current macro motion vector position calculation directioin parameter: same step 5 (1) calculates the value of motion vector directioin parameter k.

(2) judge current predicted macroblock motion vector distribution direction:

When k＜0.25, according to MV _xValue is judged: work as MV _xJudged in＞0 o'clock current macro motion vector direction be (22.5 °, 22.5 °], work as MV _xJudged in＜0 o'clock current macro motion vector direction be (157.5 °, 202.5 °];

As k＞1.5 or MV _x=0 o'clock, according to MV _yValue is judged: work as MV _yJudged in＞0 o'clock current macro motion vector direction be (67.5 °, 112.5 °], work as MV _yJudged in＜0 o'clock current macro motion vector direction be (112.5 ° ,-67.5 °];

When 0.25≤k≤1.5, according to MV _xAnd MV _yValue is judged: work as MV _x＞0 and MV _yJudged in＞0 o'clock current macro motion vector direction be (22.5 °, 67.5 °], work as MV _x＜0 and MV _yJudged in＜0 o'clock current macro motion vector direction be (157.5 ° ,-112.5 °], work as MV _x＜0 and MV _yJudged in＞0 o'clock current macro motion vector direction be (112.5 °, 157.5 °], work as MV _x＞0 and MV _yJudged in＜0 o'clock current macro motion vector direction be (67.5 ° ,-22.5 °].

Step 10 is carried out non-homogeneous multilayer hexagonal mesh search: predict the outcome according to current macro motion vector distribution in the step 9 and select the motion estimation search zone adaptively: determine the position of motion vector in 32 zones of non-homogeneous multilayer hexagonal mesh search pattern by motion vector size and motion vector direction.In the region according to 1～6 of motion vector distribution characteristic allocation search point, with the optimal match point that the searches initial ranging central point as step 11.

Step 11, expand symmetrical hexagon search: be that step-length is carried out symmetrical hexagon search with 2 pixel precisions earlier, be step-length carry out diamond search with 1 pixel precision again with optimal match point as starting point, final optimal match point is the optimal match point of current macro integer pixel motion estimation search;

Step 12, output movement estimated coding information comprises estimation time (Total ME time), code check (Bit-rate), Y-PSNR (Y-PSNR), to estimate the encryption algorithm quality.

The present invention has carried out template redesign, motion vector distribution prediction and has judged whether motion vector is 0 to select whether to skip the combination that template is searched for three kinds of technology on the advantage basis that keeps former UMHexagonS algorithm and NUMHexagonS algorithm.Whole motion estimation search main flow as shown in Figure 6, wherein motion vector direction prediction sub-process is as shown in Figure 5.Motion estimation search algorithm of the present invention had both guaranteed the accuracy of motion search, had increased substantially the real-time performance of motion-estimation encoded again.

Beneficial effect fruit of the present invention: compare with the present H.264 middle UMHexagonS algorithm that adopts, the present invention is guaranteeing under the encoded video quality prerequisite, the code check increase has been controlled in strictness and video quality descends, reduce motion estimation search effectively and counted, well improved estimation this shortcoming consuming time.

Description of drawings

Fig. 1 estimates schematic diagram for block matching motion;

Fig. 2 is the asymmetric cross multilayer of NUMHexagonS hexagon search algorithm schematic diagram;

Fig. 3 is unsymmetrical cross searching template of the present invention;

Fig. 4 is non-homogeneous hexagonal mesh search pattern of the present invention;

Fig. 5 is motion vector direction prediction sub-process figure of the present invention;

Fig. 6 is motion estimation search main flow chart of the present invention;

The rate distortion curve (HARBOUR) that Fig. 7 compares with other algorithms for motion estimation algorithm of the present invention;

The rate distortion curve (ICE) that Fig. 8 compares with other algorithms for motion estimation algorithm of the present invention.

Embodiment

Below in conjunction with drawings and Examples the present invention is done more detailed description.

In order to check the present invention to propose the validity of method, selected the cycle tests of different characteristics, as the comparatively violent video sequence ICE of motion; Comparatively mild video sequence HARBOUR moves; Video sequence MOBILE with more details smooth motion.And the UMHexagonS algorithm from the reference software JM18.4 of these three performances of motion-estimation encoded time, compression bit rate and Y-PSNR and standard H.264/AVC compares.The experiment condition configuration is as follows: at the 4G internal memory, move on the computer of 3.4GHz dominant frequency; 100 frames of encoding, frame per second 30f/s, code flow structure are IPPP, and quantization parameter QP is made as 40, and entropy is encoded to CAVLC, 5 reference frames.

Because this method is to finish at the luminance component in the video sequence, read the video sequence of one section yuv format in actual use earlier, extract its luma component information value, encoder calls the inter macroblocks fast motion estimation module of the present invention's design and finishes concrete video compression coding.

Method main flow chart of the present invention is finished following steps as shown in Figure 6 in computer:

The first step is read in the video sequence of yuv format according to encoder configuration file encoder.cfg, according to the parameter configuration encoder in the configuration file.

Second goes on foot, and reads out the luma component values of video sequence from the video file of original yuv format, takes out the luma component values that needs coded macroblocks in order.

In the 3rd step, determine initial search point: current predicted macroblock is median prediction MV _{Pred_space}, upper strata piece prediction MV _{Pred_uplayer}With time prediction MV _{Pred_ref}

Median prediction MV _{Pred_space}Utilized the correlation between adjacent macroblocks, the motion vector of current predicted macroblock can be obtained by the fast motion-vector prediction of woods around known.

Upper strata piece prediction MV _{Pred_uplayer}Utilized interframe movement to estimate the macroblock partitions feature, current predicted macroblock motion vector is obtained co-located, upper level in proper order and is twice the motion vector of sized blocks from the hierarchical search of 6 patterns.

Time prediction MV _{Pred_ref}Utilize the motion vector of current macro in different reference frames to have correlation, the current macro motion vector is predicted by a certain percentage by the motion vector of current block in the reference frame before and is obtained.

Similar motion vector MV, corresponding sad value also has very strong correlation.The rate distortion costs value of these three kinds of motion vector prediction mode points pointed is designated as space_pred respectively _Mincost, uplayer_pred _MincostAnd ref_pred _Mincost, will have the motion vector points point of minimum rate distortion costs value as initial search point, the optimum Match of this step is named a person for a particular job and is wanted the terminal point of motion vectors as next step.

In the 4th step, judge whether motion vector is 0: the motion vector that obtains the current macro optimal match point in the previous step is designated as (MV _x, MV _y), if MV _x=0 and MV _y=0, then be considered as this moment motion vector be 0, skipped for the 5th step, be that initial search point directly carried out for the 8th step with current optimal match point.

In the 5th step, current macro is carried out the motion vector direction prediction: according to next step unsymmetrical cross searching template characteristics current predicted macroblock motion vector is divided into four direction, i.e. horizontal positive direction, horizontal negative direction, vertical positive direction and vertical negative direction.Current macro motion vector coordinate (MV _x, MV _y), calculate

Value.Utilize the criterion in the step 5 (2) to judge current macro motion vector direction.

The 6th step, unsymmetrical cross searching: according to the adaptive selection direction of search of current macro motion vector distribution direction prediction result: 8 search points when being judged to be horizontal positive direction on the horizontal positive direction of search initial point, 8 search points when being judged to be horizontal negative direction on the horizontal negative direction of search initial point, 4 search points when being judged to be vertical positive direction on the vertical positive direction of search initial point, 4 search points when being judged to be vertical positive direction on the vertical positive direction of search initial point.The unsymmetrical cross searching template as shown in Figure 3.In this implementation process, search box size is 32 * 16, and step-length is 2 pixels between the adjacency search point, judges that even search needs 8 candidate search points in the horizontal direction, and search in vertical direction needs 4 candidate search points.Candidate search point is calculated the rate distortion costs value of its motion vector respectively, therefrom choose the point of rate distortion costs value minimum as the optimal match point of this step, will want the terminal point of motion vectors as next step.

In the 7th step, current macro is carried out motion vector size prediction: current predicted macroblock motion vector size is divided into motion vector is big, motion vector is medium and less three ranks of motion vector.Current predicted macroblock motion vector size Forecasting Methodology is as follows:

Determine pred according to step 7 (1) _MincostAnd RD _MincostValue, judge according to step 7 (3) the motion vector size of current predicted macroblock in step 7 (2), to have defined motion vectors parameter γ, δ, α ₁, α ₂With motion vectors bound threshold value.

The 8th step, optionally carrying out 5 * 5 pixels searches for entirely: judged current macro motion vector size in the 7th step after, only the current macro motion vector hour or motion vector be 0 o'clock just centered by current future position, in 4 * 4 zones around it, carry out the full search of 5 * 5 search points; Do not carry out 5 * 5 search entirely when motion vector size when medium or big, directly entered for the 9th step.

The 9th step, current macro is judged again whether motion vector is 0, and carry out the prediction of motion vector distribution direction prediction and motion vector size: the optimal match point in the 8th step is predicted again, judged that wherein whether motion vector is 0 to go on foot identical with motion vector size Forecasting Methodology and above-mentioned the 4th step and the 7th.Motion vector is 0 if predict the outcome, and then skips for the tenth step, directly carries out for the 11 step.If motion vector is not 0, then obtains the current macro motion vector distribution by motion vector direction prediction and the prediction of motion vector size and predict the outcome.Motion vector distribution direction prediction method is as follows in this step:

Calculate the value of motion vector directioin parameter k according to step 5 (1).Carrying out current macro motion vector distribution direction according to step 9 (2) judges.

In the tenth step, non-homogeneous multilayer hexagonal mesh is searched for: predict the outcome according to current macro motion vector distribution in the step 9 and select the motion estimation search zone adaptively: determine the position of motion vector in 32 zones of non-homogeneous multilayer hexagonal mesh search pattern by motion vector size and motion vector direction.Non-homogeneous multilayer hexagonal mesh search pattern as shown in Figure 4, in the region of search according to the motion vector distribution characteristic allocation 1～6 the search point.With the optimal match point that searches in this step initial ranging central point as step 11.

The 11 step, expand symmetrical hexagon search: be that step-length is carried out symmetrical hexagon search with 2 pixel precisions earlier, be step-length carry out diamond search with 1 pixel precision again with optimal match point as starting point, final optimal match point is the optimal match point of current macro integer pixel motion estimation search.

The 12 step finished motion estimation search, preserved output movement estimated coding information, comprised estimation time, code check, Y-PSNR, to estimate the encryption algorithm quality.

Experimental result is as shown in table 1.As can be seen from Table 1, the UMHexagonS algorithm among the reference software JM18.4 of method of the present invention and standard is H.264/AVC compared, and Y-PSNR slightly reduces, and on average reduces by 0.05%, and video quality slightly descends in the human vision scope; Strict control code check on average increases by 0.35%, has kept the advantage of primary standard algorithm high compression ratio; Shortened the motion-estimation encoded time effectively, on average saved for 21.71% motion-estimation encoded time.Fig. 7 represents to represent the mild video sequence (HARBOUR) of motion respectively with Fig. 8 and moves violent video sequence (ICE) under different quantization steps (QP) situation, motion estimation algorithm of the present invention and UMHexagonS algorithm rate distortion curve comparative result with representative.Show that by table 1 and Fig. 7,8 the present invention has improved the motion-estimation encoded real-time performance well, and kept the H.264 advantage of the low high video quality of code check.Algorithm optimization effect of the present invention is obvious and stable, effectively reduces the complexity of motion estimation algorithm architecture.

UMHexagonS algorithm motion-estimation encoded performance relatively among table 1 motion estimation algorithm of the present invention and the JM18.4

Claims (5)

1. quick video motion estimated coding method based on motion vector distribution prediction, it is characterized in that this method selects the template region of search adaptively according to current macro motion vector distribution prediction, whether be 0 to predict to motion vector on a large scale before the template search, add the premature termination criterion, and then realization specifically may further comprise the steps the coding quick movement estimation video of inter macroblocks:

Step 1 is extracted the monochrome information of current predicted macro block as coded object from current video frame;

Step 2 is determined twofold motion estimating searching template, carries out search point and distributes;

Step 3, determine the initial search point of high precision:

Adjudicate foundation by Lagrangian rate-distortion optimization (RDO-Rate Distortion Optimization) function as estimation, optimum Match prediction piece and optimal motion vector on the selection rate distortion sense make the Bit Allocation in Discrete minimum of motion vector and residual coding; Utilize Lagrangian rate distortion criterion to select the optimal motion vector problem can be described as:

J _motion(mv,ref|λ _motion)＝SAD[s,r(ref,mv)]+λ _motion[R(mv-pred)+R(ref)]

Wherein, J _MotionRate distortion costs value for the motion vector of current prediction

S is the current macro block pixels value, mv is current motion vector, pred is motion vectors, ref is reference frame, r (ref, mv) be the pixel value of reference macroblock, R is that motion vector carries out the bit number that differential coding consumes, and comprises the differential coding bit number of motion vector and its predicted value and the number of coded bits of reference frame; λ _MotionBe Lagrange multiplier, SAD(Sum of Absolute Difference) be between current block and reference block pixel absolute error and:

$\mathrm{SAD}[s,r(\mathrm{ref},\mathrm{mv})]=\underset{x=1}{\overset{B_1}{\mathrm{\Σ}}}\underset{y=1}{\overset{B_2}{\mathrm{\Σ}}}|s(x,y)-r(x-m_x,u-m_y)|$

Wherein, B ₁And B ₂Horizontal pixel and the vertical pixel of representing piece respectively, according to different inter-frame forecast modes, its value can be 16,8, and 4; (x y) is the current macro block pixels value to s; (x y) is the pixel value of reference macroblock, m to r _xAnd m _yThe displacement of representing level and vertical direction respectively;

Utilize Lagrangian rate distortion criterion to select the problem of optimization model to be expressed as:

J _mode(s,c,MODE|λ _mode)＝SSD(s,c,MODE|QP)+λ _mode×R(s,c,MODE|QP) （1）

Wherein, MODE represents a kind of interframe encoding mode of current macro, J _Mode(s, c, MODE| λ _Mode) rate distortion costs value under the expression MODE pattern S is original vision signal, and c is the reconstructed video signal behind the employing MODE pattern-coding, λ _ModeBe Lagrange multiplier, R (s, c, MODE|QP) be the total number of bits that comprise macro block header, motion vector and all DCT block messages relevant with MODE and quantization parameter, QP is the coded quantization step-length, SSD (s, c, MODE|QP) (Sum of Square Difference) is the difference of two squares sum between primary signal and the reconstruction signal, that is:

$\mathrm{SSD}(s,c,\mathrm{MODE}|\mathrm{QP})=\underset{i=1,j=1}{\overset{B1,B2}{\mathrm{\Σ}}}{[S_Y(x,y)-C_Y(x,y,\mathrm{MODE}|\mathrm{QP})]}^2$

$+\underset{i=1,j=1}{\overset{B1,B2}{\mathrm{\Σ}}}{[S_U(x,y)-C_U(x,y,\mathrm{MODE}|\mathrm{QP})]}^2+\underset{i=1,j=1}{\overset{B1,B2}{\mathrm{\Σ}}}{[S_V(x,y)-C_V(x,y,\mathrm{MODE}|\mathrm{QP})]}^2$

Wherein, B ₁And B ₂Represent horizontal pixel and the vertical pixel of piece respectively, its value can be 16,8, and 4; S _Y(x y) is the value of source macro block brightness information, C _Y(MODE|QP) value of the monochrome information of macro block, S are rebuild in expression for x, y _U, S _VAnd C _U, C _VThe value of representing corresponding colour difference information respectively;

Specifically may further comprise the steps:

(1) adopt multiple prediction mode to carry out the starting point prediction:

1) carries out the median prediction MV of spatial domain _{Pred_space}: utilize the correlation between adjacent macroblocks, the motion vector of current predicted macroblock is obtained by the motion-vector prediction of adjacent piece around known;

2) carry out the upper strata piece prediction MV that many sized blocks are divided _{Pred_uplayer}: utilize interframe movement to estimate the macroblock partitions feature, current predicted macroblock motion vector is obtained co-located, upper level in proper order and is twice the motion vector of sized blocks from the hierarchical search of 7 patterns;

3) carry out the reference frame motion-vector prediction MV of time-domain _{Pred_ref}: utilize the motion vector of current macro in different reference frames to have correlation, the current macro motion vector is predicted by a certain percentage by the motion vector of current block in the reference frame before and is obtained;

(2) the rate distortion costs value with above-mentioned three kinds of motion vector prediction mode points pointed is designated as space_pred respectively _Mincost, uplayer_pred _MincostAnd ref_pred _Mincost, will have the motion vector points point of minimum rate distortion costs value as initial search point, the optimal match point of this step is wanted the terminal point of motion vectors as next step;

Step 4, judge whether motion vector is 0:

The motion vector that obtains the current macro optimal match point in the previous step is designated as (MV _x, MV _y), if MV _x=0 and MV _y=0, then be considered as this moment motion vector be 0, skips steps five is directly carried out step 8;

Step 5, the prediction current macro is carried out the motion vector direction: according to next step unsymmetrical cross searching template characteristics current predicted macroblock motion vector is divided into four direction, i.e. horizontal positive direction, horizontal negative direction, vertical positive direction and vertical negative direction;

Step 6, carry out unsymmetrical cross searching: the vertical search scope is half of horizon scan scope, step-length is two pixels between the adjacency search point; Select the direction of search adaptively according to current macro motion vector direction prediction result in the 5th step: when being judged to be horizontal positive direction, 8 search points on the horizontal positive direction of search initial point; When being judged to be horizontal negative direction, 8 search points on the horizontal negative direction of search initial point; When being judged to be vertical positive direction, 4 search points on the vertical positive direction of search initial point; When being judged to be vertical negative direction, 4 search points on the vertical negative direction of search initial point; The optimum Match of determining in this step is named a person for a particular job and is wanted the terminal point of motion vectors as next step;

Step 7 is divided into current predicted macroblock motion vector size that motion vector is big, motion vector is medium and less three ranks of motion vector, judges current macro motion vector size;

Step 8 is optionally carried out 5 * 5 pixels and is searched for entirely: optionally carries out 5 * 5 pixels according to prediction current macro motion vector size result in the step 7 and search for entirely; Only motion vector hour or motion vector be 0 o'clock just centered by current future position, in 4 * 4 zones around it, carry out the full search of 5 * 5 search points; Otherwise do not carry out 5 * 5 search entirely, directly enter step 9;

Step 9, current macro is judged again whether motion vector is 0, and carry out the prediction of motion vector direction prediction and motion vector size: the optimal match point in the step 8 is predicted again, judged that wherein whether motion vector is 0 identical with above-mentioned steps four and step 7 with motion vector size Forecasting Methodology; If motion vector is 0, then skips steps ten, directly carry out step 11; If motion vector is not 0, then obtains the current macro motion vector distribution by motion vector direction prediction and the prediction of motion vector size and predict the outcome; According to next step non-homogeneous multilayer hexagonal mesh search pattern characteristics current predicted macroblock motion vector is divided into eight directions:

(-22.5°,22.5°]，(157.5°,202.5°]，(22.5°,67.5°]，(-157.5°,-112.5°]，

(67.5°,112.5°],(-112.5°,-67.5°],(112.5°,157.5°],(-67.5°,-22.5°]；

Step 10 is carried out non-homogeneous multilayer hexagonal mesh search: predict the outcome according to current macro motion vector distribution in the step 9 and select the motion estimation search zone adaptively: determine the position of motion vector in 32 zones of non-homogeneous multilayer hexagonal mesh search pattern by motion vector size and motion vector direction; In the region according to 1～6 of motion vector distribution characteristic allocation search point, with the optimal match point that the searches initial ranging central point as step 11;

Step 11, expand symmetrical hexagon search: be that step-length is carried out symmetrical hexagon search with 2 pixel precisions earlier, be step-length carry out diamond search with 1 pixel precision again with optimal match point as starting point, final optimal match point is the optimal match point of current macro integer pixel motion estimation search;

Step 12, output movement estimated coding information comprises estimation time (Total ME time), code check (Bit-rate), Y-PSNR (Y-PSNR), to estimate the encryption algorithm quality.

2. the quick video motion estimated coding method based on the motion vector distribution prediction according to claim 1 is characterized in that, step 2 determines that the method for estimation template distribution search point is as follows:

(1) unsymmetrical cross searching template: formed by 4 regions of search, distribute 4 search points vertically respectively, distribute 8 search points about level respectively;

(2) non-homogeneous hexagonal mesh search pattern: constitute 32 regions of search jointly by 4 layers and 8 directions, the non-homogeneous distribution of counting is carried out in each zone, increase along with search radius, the expansion of hunting zone, distribute search point to increase progressively from the inside to the outside, have 62 of search points in the horizontal direction ± 45 in ° zone, have 22 of search points in the vertical direction ± 45 in ° zone.

3. the quick video motion estimated coding method based on the motion vector distribution prediction according to claim 1 is characterized in that, the method that step 5 prediction current macro is carried out the motion vector direction may further comprise the steps:

(1) according to current macro motion vector position calculation directioin parameter:

Current macro motion vector coordinate (MV _x, MV _y), the motion vector direction is with its direction vector MV=(MV _x, MV _y) expression, calculate Value;

(2) judge current predicted macroblock motion vector direction:

When k＜1, according to MV _xState is judged: work as MV _xJudged that current macro motion vector direction was horizontal positive direction at＞0 o'clock, work as MV _xJudged that current macro motion vector direction was horizontal negative direction at＜0 o'clock;

Work as k〉1 or MV _x=0 o'clock, according to MV _yState is judged: work as MV _yJudged that current macro motion vector direction was vertical positive direction at＞0 o'clock, work as MV _yJudged that current macro motion vector direction was vertical negative direction at＜0 o'clock.

4. the quick video motion estimated coding method based on the motion vector distribution prediction according to claim 1 is characterized in that, step 7 judges that the method for current macro motion vector size is as follows:

(1) determines pred according to the prediction mode of initial search point _MincostAnd RD _MincostValue:

If in step 3, adopt the time prediction mode, then: pred _Mincost=ref_pred _Mincost

If in step 3, adopt the spatial prediction mode, then: pred _Mincost=space_pred _Mincost

If in step 3, adopt upper strata piece prediction mode, then: pred _Mincost=uplayer_pred _Mincost

RD _MincostFor by calculating the minimum rate distortion costs value of current predicted macroblock under the MODE pattern in the formula (1);

(2) calculate the motion vectors parameter:

The motion vectors lower threshold is:

(1+γ)pred _mincost

The motion vectors upper limit threshold is:

(1+δ)pred _mincost

The motion vectors factor is:

$\mathrm{\γ}=\frac{\mathrm{Bsize}\left[\mathrm{blocktype}\right]}{{{\mathrm{pred}}_{\min \cos t}}^2}-{\mathrm{\α}}_1\left[\mathrm{blocktype}\right]$

$\mathrm{\δ}=\frac{\mathrm{Bsize}\left[\mathrm{blocktype}\right]}{{{\mathrm{pred}}_{\min \cos t}}^2}-{\mathrm{\α}}_2\text{[blocktype]}$

Wherein, Bsize[blocktype] size of the current predicted macroblock of expression, its value can be 16,8,4,

α ₁[blocktype]＝[-0.23,-0.23,-0.23,-0.25,-0.27,-0.27,-0.28]

α ₂[blocktype]＝[-2.39,-2.40,-2.40,-2.41,-2.45,-2.45,-2.48]

(3) judge current predicted macroblock motion vector size:

Work as RD _Mincost≤ (1+ γ) pred _Mincost, when namely minimum rate distortion costs value is less than the motion vectors lower threshold, judge that the current macro motion vector is less, namely movement degree is mild;

As (1+ γ) pred _Mincost＜RD _Mincost＜(1+ δ) pred _Mincost, when namely minimum rate distortion costs value is between motion vectors lower threshold and motion vector upper limit threshold, judge that the current macro motion vector is medium, namely movement degree is medium;

Work as RD _Mincost〉=(1+ δ) pred _Mincost, when namely minimum rate distortion costs value is greater than the motion vectors upper limit threshold, judge that the current macro motion vector is bigger, namely movement degree is more violent.

5. the quick video motion estimated coding method based on motion vector distribution prediction according to claim 1 is characterized in that the method for step 9 motion vectors direction is as follows:

(1) according to current macro motion vector position calculation directioin parameter: same step 5 (1) calculates the value of motion vector directioin parameter k;

(2) judge current predicted macroblock motion vector distribution direction:

When k＜0.25, according to MV _xValue is judged: work as MV _x＞0 o'clock, judge current macro motion vector direction be (22.5 °, 22.5 °]; Work as MV _x＜0 o'clock, judge current macro motion vector direction be (157.5 °, 202.5 °];

As k＞1.5 or MV _x=0 o'clock, according to MV _yValue is judged: work as MV _y＞0 o'clock, judge current macro motion vector direction be (67.5 °, 112.5 °]; Work as MV _y＜0 o'clock, judge current macro motion vector direction be (112.5 ° ,-67.5 °];

When 0.25≤k≤1.5, according to MV _xAnd MV _yValue is judged: work as MV _x＞0 and MV _y＞0 o'clock, judge current macro motion vector direction be (22.5 °, 67.5 °]; Work as MV _x＜0 and MV _y＜0 o'clock, judge current macro motion vector direction be (157.5 ° ,-112.5 °]; Work as MV _x＜0 and MV _yJudged in＞0 o'clock current macro motion vector direction be (112.5 °, 157.5 °]; Work as MV _x＞0 and MV _y＜0 o'clock, judge current macro motion vector direction be (67.5 ° ,-22.5 °].

Images