CN103384324A - Quick sub pixel motion estimation method for AVS-M video coding - Google Patents
Quick sub pixel motion estimation method for AVS-M video coding Download PDFInfo
- Publication number
- CN103384324A CN103384324A CN2013100190213A CN201310019021A CN103384324A CN 103384324 A CN103384324 A CN 103384324A CN 2013100190213 A CN2013100190213 A CN 2013100190213A CN 201310019021 A CN201310019021 A CN 201310019021A CN 103384324 A CN103384324 A CN 103384324A
- Authority
- CN
- China
- Prior art keywords
- search
- pixel
- point
- motion
- optimum
- 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
Images
Abstract
The invention provides a quick sub pixel motion estimation method for AVS-M video coding. As shown in a figure, when an integer pixel portion for forecasting a motion vector is equal to the optimal vector obtained by integer pixel motion search, the optimal appropriate vector of a current piece and a forecasting sub pixel motion vector with the maximum matching error in motion vectors of an upper layer mode are used as initial search points. The step of small rhombus search is carried out, and the position of the optimal point and the position of the suboptimum point are stored. If the optimal point is the center of a small rhombus, search is over. If not, new small rhombus search is carried out, self-adaptive changing threshold value judgment is added into a search process, and unnecessary search steps are removed. When the forecasting vector is not equal to the optimal motion vector searched by integer pixel, sub pixel is utilized to carry out fine search. To video sequences with different motion characteristics, algorithm can guarantee image quality and coding efficiency to be unchanged basically to effectively reduce calculated amount of sub pixel motion estimation and improve coding speed.
Description
Technical field
The present invention relates to the technical field of video coding in the signal processing, be specifically related to a kind of fast sub-picture element movement estimating method of AVS-M Video coding.Described " AVS-M " full name is " digital audio/video mobile multimedia national standard ".
Background technology
Since C.E.Shannon in 1948 proposed source coding theory, people had carried out a large amount of research to Image Compression Coding Technology.Through the research and development of six more than ten years, a lot of technology and method have appearred.AVS(Audio Video Coding Standard) be that China independently formulates, have the Audio Video coding Standard of independent intellectual property right, AVS-M is that China is towards the video encoding and decoding standard of mobile communication.
Estimation (Motion Estimation, ME) is the key component of AVS-M Video coding, is also simultaneously amount of calculation maximum, maximum part consuming time in whole encoder.Motion estimation module is estimated by whole pixel motion and sub-pix estimation two parts form, estimate the development of fast algorithm along with whole pixel motion, the search point of whole pixel significantly reduces, generally can be down to below 10 points, the proportion that causes the amount of calculation of sub-pix estimation to account for whole motion estimation process amount of calculation increases.Therefore, the amount of calculation that how to reduce the sub-pix estimation has just become to realize a study hotspot of real-time video coding.
Present many scholars have carried out a large amount of research to the sub-pix motion estimation algorithm, and wherein typical algorithm has: center-biased sub-pix searching algorithm CBFPS(Center Biased Fractional Pixel Search); Parabola prediction algorithm (Parabolic Prediction based Fractional Pixel Search, PPFPS) based on sub-pix; Sub-pix adaptive range searching algorithm (Adaptive Range Fractional Pixel Search, ARFPS).These algorithms have been obtained effect preferably in some characteristic image, but picture material varies, and when the pixel value regularity of distribution and hypotheses were not inconsistent, the algorithmic match effect sharply descended.
The CBFPS algorithm, at first utilize the median prediction of adjacent block motion vector to go out current block at the motion vector of sub-pixel location, then comparison prediction sub-pix motion vector and the search initial point matching error, get matching error than point as the search starting point, utilize little diamond search template (SDSP) to do fine search at 1/4 pixel precision.Compare with full search, algorithm can reduce by 53% amount of calculation, yet for the large-size piece, the sub-pix motion vector of prediction is often not accurate enough, easily causes the larger decline of picture quality.
Many fast motion estimation algorithms utilize the motion vector of adjacent block to predict the search starting point of current block, with the search point that reduces the piece coupling with prevent from being absorbed in local minimum point.Motion vectors comprises whole pixel and sub-pix two parts information, if the whole pixel portion of motion vectors equals the optimum vector that whole pixel motion search obtains, the sub-pix of predictive vector part is probably near global optimum's point so, and we can begin search as the search starting point with the sub-pix motion vector of prediction.Otherwise not accurate enough prediction sub-pix motion vector can make matching effect sharply descend, and the present invention is directed to this situation and proposes a kind of search strategy, when guaranteeing picture quality, can effectively reduce search point.
Summary of the invention
For the video encoding and decoding standard AVS-M needs more efficient fast sub-picture element motion estimation algorithm of China towards mobile communication, in order to solve the deficiencies in the prior art, the present invention discloses a kind of AVS-M Video coding fast sub-picture element movement estimating method.The method can reduce the amount of calculation of sub-pix estimation effectively when guaranteeing that picture quality and code efficiency are substantially constant, improve coding rate.
For achieving the above object, the present invention adopts following sub-picture element movement estimating method.
Step1 calculates the matching error of best whole pixel
J MOTIONIf the whole pixel portion of predictive vector equals the optimal motion vector that whole pixel search obtains, and forwards Step2 to, otherwise, forward Step5 to.
Step2 calculates
MVP ClosestWith
MVP Up-layer J MOTION, select to have minimum match error Min
J MOTIONPoint, if Min
J MOTION<
TH i , forward Step6 to, otherwise, forward Step3 to.
Step3 is with Min
J MOTIONDo a little diamond search of step centered by point, preserve position optimum and time advantage, if optimum point is the center of little rhombus, forward step6 to, otherwise, forward Step4 to.
Step4 does new little diamond search, if center or little diamond search step number that optimum point is new little rhombus reach 3, forwards Step6 to, otherwise, repeat Step4.
Step5 does the sub-pix fine search.
It is the motion vector that final inferior number pixel is searched for that Step6 selects the motion vector of current optimum point, and inferior number pixel search finishes.
Compared with the prior art, advantage of the present invention and good effect are: for the video sequence of different motion feature, this algorithm is when guaranteeing that picture quality and code efficiency are substantially constant, compare with full-search algorithm, search point reduces 65.25%~87.69%, can effectively reduce the amount of calculation of sub-pix estimation, improve coding rate.And algorithm do not need to test the matching error of point around optimum whole pixel motion vector, can be used in conjunction with any whole pixel searching algorithm.
Affect mainly containing of motion estimation algorithm performance: the search starting point, search abort criterion and search pattern, the below describes in detail this three part.
Determining of search starting point of the present invention.
The best of adjacent block is near vector predictionThe motion vector of adjacent block and the motion vector of current block have very strong correlation, AVS-M utilize this correlation to the motion vector of present encoding piece and its best near vector
MVP ClosestDifferential coding is to reduce the required bit number of encoding motion vector.The present invention chooses
MVP ClosestA predictive vector as current block sub-pix motion vector is referred to as the best near vector.
The best is near vector
MVP ClosestBe defined as follows.
In accompanying drawing 2, piece A, B, C, D are respectively left, top, upper right side and the upper left adjacent blocks of current block E, and the distance of the motion vector of definition block A, B, C is:
Get
VAB,
VBC,
VCAIn maximum, if maximum is
VAB, the motion vector of getting the C piece as the best near vector, if maximum is
VBC, the motion vector of getting the A piece as the best near vector, otherwise the motion vector of getting the B piece as the best near vector.
The motion-vector prediction of upper mode7 kinds of inter-frame forecast modes are arranged in AVS-M, as shown in Figure 3, there is very strong correlation in the motion vector between different size block, therefore, (mode 1 can first to search for large piece, 2,3), then search for little piece ( mode 4,5,6,7), so just can predict the wherein motion vector of fritter with the motion vector of the bulk of finding out.The levels that the present invention defines 7 kinds of inter-frame modes is closed and to be: pattern 1 is the upper mode of pattern 2 and 3, and pattern 2 is upper modes of pattern 4, and pattern 4 is upper modes of pattern 5 and 6, and pattern 5 is upper modes of mode 7, and pattern 1 does not have the upper strata.Get the motion vector of upper mode
MVP Up-layerPredictive vector as present mode.
Determining of initial ranging starting point.
Statistics such as the table 1 of prediction sub-pix motion vector deviation.
As can be seen from Table 1, for the Foreman sequence, the best is respectively 41.51% and 47.33% near the identical probability of optimum movement vector that predictive vector and upper strata predictive vector and whole pixel search obtain, if utilize simultaneously this two kinds of predictive vectors, identical probability is increased to 62.91%, two kinds of predictive vectors of algorithm picks of the present invention only adopt the accuracy of median prediction vector high a lot of than CBFPS.Although with predictive vector as final motion vector, this probability is also enough not high, if but with the best near predictive vector and upper strata predictive vector as initial search point, do a little diamond search of step, just can obtain 84.75% optimum movement vector, do three little diamond search of step, accuracy rate namely reaches 97.44%.Have the sequence of less motion feature for Akiyo etc., accuracy can reach 99.79%.Therefore, the present invention get have minimum match error prediction sub-pix motion vector as initial search point.
The statistics of table 1 prediction sub-pix motion vector deviation
The search abort criterion.
When the whole pixel portion of motion vectors equals optimum vector that whole pixel motion search obtains, get have minimum match error prediction sub-pix motion vector as initial search point, do three little diamond search of step and obtain final motion vector.
As can also be seen from Table 1, for Akiyo, News, the Foreman sequence, in all blocks to be encoded, final sub-pix motion vector has respectively 83.73%, 81.79% is identical with predictive vector with 62.91%, if can dope these encoding blocks, just can skip the SDSP search procedure, save a large amount of unnecessary calculating.If predictive vector is identical with final sub-pix motion vector, predictive vector will have minimum matching error.
Matching criterior commonly used in motion estimation algorithm has three kinds, namely minimum absolute difference (
MAD), least mean-square error (
MSE) and Normalized Cross Correlation Function (
NCCF).In estimation, matching criterior is not very large to the Accuracy of coupling, due to
MADCriterion need not done multiplying, and realization is simple, convenient, so use at most common the use
SADReplace
MAD,
SAD(Sum of Absolute Difference) is the absolute error summation.Be defined as follows:
(5)。
AVS-M has adopted multimodal estimation, and under different predictive modes, the number of coded bits that its motion vector consumes is different, and the required number of coded bits of error coding after prediction is also different.So AVS-M considers predicated error summation and the required bit number of final coding, adopt rate-distortion optimization criterion (RDO) judgement optimal motion vector in motion estimation and compensation.Optimum movement vector search in AVS-M is exactly to seek to make formula (6) be the motion vector of minimum value in the hunting zone:
In formula
, be motion vector;
Motion vector for prediction;
It is Lagrange (Langrange) multiplier factor;
R(m-p) be look-up table about movable information;
SADBe absolute error and (residual signals) between source piece and reference block.
AVS-M is with rate distortion costs
J MOTIONAs the criterion of optimal motion vector, so will investigate
J MOTIONCorrelation between adjacent block.
The threshold adaptive that the present invention proposes withdraws from the core concept of algorithm in advance with predictive vector
J MOTIONDecide the motion search process with given threshold ratio.Obtain the candidate value of several groups of threshold values by test, then carry out computing with every group of threshold value, more every group of resulting Y-PSNR of threshold value, bit rate and search point are compared, finally obtain the optimal threshold that needs.
Statistical analysis to the various video sequence shows, same position piece between consecutive frame
J MOTIONStill have very strong correlation, ubiquity certain proportionate relationship, and its mean ratio is between 0.96~1.06.
Algorithm of the present invention has adopted following method definite threshold: with reference frame same position piece
J MOTIONMultiply by certain coefficient conduct
J MOTIONThreshold value.This method is not subjected to the impact of video motion type, has stronger adaptivity, and can determine accurately that appropriate threshold value is ended search procedure in good time, helps to obtain Search Results preferably with less search cost.For this reason, 7 kinds of piece sizes are arranged adaptive threshold
TH i :
In formula
Prev_J MOTIONBe the mean value of same block size sub-pix minimum match error in reference frame,
αBe regulatory factor, it has extremely important effect in compromise reconstructed image quality and search speed.
The abort criterion of algorithm is: if
, show that the sub-pix motion vectors is enough accurate, stops search procedure immediately.
Search pattern.
New little diamond search pattern of the present inventionWhen the whole pixel portion of motion vectors equals optimum vector that whole pixel motion search obtains, get have minimum match error prediction sub-pix motion vector as initial search point, the present invention adopts a kind of new little diamond search pattern to obtain final motion vector.
The new little diamond search process that the present invention adopts as shown in Figure 4.In Fig. 4,1A, 1B are optimum point and the inferior advantage that the little diamond search of the first step obtains, 2,2A is the point of the little diamond search of second step, and 2A, 1A are optimum point and the inferior advantage that the little diamond search of second step obtains, 3 is the point of the 3rd little diamond search of step.First carry out a little diamond search of step, if optimum point is not the central point of little rhombus, when the little diamond search of second step, the search point identical with inferior advantage position with optimum point in the little diamond search of the first step only.The rest may be inferred in the 3rd little diamond search of the step search point identical with inferior advantage position with optimum point in the little rhombus of second step only.When the point of minimum match error was positioned at the center of rhombus or search step number and reached for three steps, search stopped.
Sub-pix fine search pattern of the present inventionIf the whole pixel portion of predictive vector is different from the optimal motion vector that whole pixel motion search obtains, just can not use the sub-pix part of predictive vector in the sub-pix search, because not accurate enough prediction sub-pix motion vector can cause the larger decline of picture quality.The present invention adopts the sub-pix fine search, guarantees the accuracy of sub-pix search, and determines the position of next step search point, the little point that excludes the possibility, thereby acceleration search process according to the Search Results of previous step.
Three kinds of situations of sub-pix fine search of the present invention are as shown in Figure 5: in 1/2 pixel precision search, take the whole pixel motion vector of optimum as search center, 1/2 pixel is unit, do a little diamond search of step, preserve optimum 1/2 pixel and suboptimum 1/2 pixel position and matching error, 1/2 pixel of testing according to other needs of determining positions of optimum and suboptimum pixel.As shown in Fig. 5 (a), if optimum 1/2 pixel and suboptimum 1/2 pixel 2A, 2B is adjacent and be not the search starting point, so only needing increases test a bit again, as 1/2 pixel 3 in figure.In Fig. 5 (b), if having one to be the search starting point in optimum or suboptimum l/2 pixel 2A or 2B, need so to increase by two adjacent 1/2 pixels 3 of test optimum point 2A.In Fig. 5 (c), if optimum and suboptimum 1/2 pixel is symmetrical and be not the search starting point, need to be tested remaining all 1/2 pixels 3.After complete 1/2 pixel motion vector of above-mentioned decision search, with the minimum match error o'clock search center as 1/4 pixel precision, 1/4 pixel is unit, does a little diamond search of step.
Description of drawings
Fig. 1 is AVS-M Video coding fast sub-picture element motion estimation algorithm flow chart of the present invention.
Fig. 2 is that the best of the present invention is near vector prediction figure.
Fig. 3 is 7 kinds of inter-frame forecast modes of AVS-M.
Fig. 4 is new little diamond search process schematic diagram of the present invention.
Fig. 5 is sub-pix fine search schematic diagram of the present invention.
Embodiment
AVS-M Video coding fast sub-picture element motion estimation algorithm flow process of the present invention as shown in Figure 1, the key step of algorithm is as follows.
Step1 calculates the matching error of best whole pixel
J MOTIONIf the whole pixel portion of predictive vector equals the optimal motion vector that whole pixel search obtains, forward Step2 to, otherwise, forward Step5 to.
Step2 calculates
MVP ClosestWith
MVP Up-layer J MOTION, select to have minimum match error Min
J MOTIONPoint, if Min
J MOTION<
TH i , forward Step6 to.Otherwise, forward Step3 to.
Step3 is with Min
J MOTIONDo a little diamond search of step centered by point, preserve position optimum and time advantage, if optimum point is the center of little rhombus, forward step6 to, otherwise, forward Step4 to.
Step4 does new little diamond search shown in Figure 4, the optimum point that obtains with the little diamond search of back is the center of new little diamond search, search in new little rhombus optimum with back and time identical point in advantage position, preserves optimum in new little diamond search and the position of inferior advantage.If center or little diamond search step number that optimum point is new little rhombus reach 3, forward Step6 to, otherwise, repeat Step4.
Step5 does sub-pix fine search shown in Figure 5.In 1/2 pixel precision search, take the whole pixel motion vector of optimum as search center, 1/2 pixel is unit, do a little diamond search of step, preserve optimum 1/2 pixel and suboptimum 1/2 pixel position and matching error, 1/2 pixel of testing according to other needs of determining positions of optimum and suboptimum pixel.If optimum 1/2 pixel is adjacent with suboptimum 1/2 pixel and be not the search starting point, so only need increase again 1/2 pixel between consecutive points; If have one to be the search starting point in optimum or suboptimum l/2 pixel, need so to increase by two adjacent 1/2 pixels of test optimum point; If optimum and suboptimum 1/2 pixel is symmetrical and be not the search starting point, need to be tested remaining all 1/2 pixels; After complete 1/2 pixel motion vector of above-mentioned decision search, with the minimum match error o'clock search center as 1/4 pixel precision, 1/4 pixel is unit, does a little diamond search of step.
It is the motion vector that final inferior number pixel is searched for that Step6 selects the motion vector of current optimum point, and inferior number pixel search finishes.
Algorithm of the present invention is inserted AVS-M Knowledge Verification Model WM3.3a, chooses 6 QCIF(174 with different motion feature * 144) standard video sequence tests algorithm performance.In 6 sequences, " Akiyo, Silent " has static background and less motion, and " News, Carphone " movement degree is placed in the middle, and " Foreman, Mobile " has stronger motion and mobile background.These video sequences have been contained various types of video sequence characteristics substantially, therefore can think that test result is of universal significance.The encoder major parameter comprises: the hunting zone is 16; Reference frame is 1; Allow RDO; The Hadamard conversion is opened.Video sequence respectively comprises 100 frames, and frame per second is that 30. quantization parameters are 28, and first frame is encoded to the I frame, and all the other frames are encoded to the P frame,
α=0.95.Test platform: CPU is Intel Pentium4 3.0G, in save as DDR512MB, operating system is Windows XP.As shown in table 2 with HPFS algorithm, CBFPS algorithm and ARFPS algorithm test result under the same conditions.
The comparison of table 2 algorithm of the present invention and other algorithm
As can be seen from Table 2, compare with the sub-pix full-search algorithm, the search point of algorithm of the present invention has reduced 65.25%~87.68%, the average sub-pix estimation of every frame time decreased 54.53%~79.69%, and mean P SNR descends and to be no more than 0.01dB.Compare with the CBFPS algorithm, search point has reduced 43.67%~67.38%, the average sub-pix estimation of every frame time decreased 35.59%~67.51%, this is because the CBFPS algorithm only adopts fast algorithm to the encoding block of 8 * 8 following piece sizes, and relatively large (16 * 16,16 * 8,8 * 16) are still used full search.The present invention uses fast algorithm to whole encoding blocks, and this is that the accuracy of predictive vector is higher than CBFPS algorithm, can be applied in larger encoding block because algorithm of the present invention has adopted more predictive vector.Compare with the ARFPS algorithm, search point has reduced 38.50%~62.33%, the average sub-pix estimation of every frame (ME) time decreased 28.71%~62.17%, mean P SNR is substantially suitable, simultaneously the variation of algorithm coding code check of the present invention is also very small.
Should be understood that; above-mentioned explanation is not to be limitation of the present invention, and the present invention also is not limited in above-mentioned giving an example, the modification that those skilled in the art make in essential scope of the present invention; distortion, interpolation or replacement also should belong to protection scope of the present invention.
Claims (4)
1. AVS-M Video coding fast sub-picture element movement estimating method, its feature comprises the following steps:
Step1 calculates the matching error of best whole pixel
J MOTIONIf the whole pixel portion of predictive vector equals the optimal motion vector that whole pixel search obtains, and forwards Step2 to, otherwise, forward Step5 to;
Step2 calculates
MVP ClosestWith
MVP Up-layer J MOTION, select to have minimum match error Min
J MOTIONPoint, if Min
J MOTION<
TH i , forward Step6 to, otherwise, forward Step3 to;
Step3 is with Min
J MOTIONDo a little diamond search of step centered by point, preserve position optimum and time advantage, if optimum point is the center of little rhombus, forward step6 to, otherwise, forward Step4 to;
Step4 does new little diamond search, if center or little diamond search step number that optimum point is new little rhombus reach 3, forwards Step6 to, otherwise, repeat Step4;
Step5 does the sub-pix fine search;
It is the motion vector that final inferior number pixel is searched for that Step6 selects the motion vector of current optimum point, and inferior number pixel search finishes.
2. a kind of AVS-M Video coding fast sub-picture element movement estimating method according to claim 1 is characterized in that in Step2:
(1)
MVP ClosestA predictive vector for current block sub-pix motion vector is referred to as the best near vector,
MVP Up-layerBe the motion vector of upper mode, comprise the piece of former frame same position and left, top and the top-right motion vector of current block
MVP 0,
MVP L,
MVP U,
MVP UR
(2) adaptive change threshold value:
Prev_ in formula
J MOTIONBe the mean value of same block size sub-pix minimum match error in reference frame, α is regulatory factor.
3. a kind of AVS-M Video coding fast sub-picture element movement estimating method according to claim 1, it is characterized in that little diamond search pattern new in Step4 is: first carry out a little diamond search of step, if optimum point is not the central point of little rhombus, when the little diamond search of second step, the search point identical with inferior advantage position with optimum point in the little diamond search of the first step only, the rest may be inferred in the 3rd little diamond search of the step search point identical with inferior advantage position with optimum point in the little rhombus of second step only, when the point of minimum match error is positioned at the center of rhombus or search step number and reached for three steps, search stops.
4. at a kind of AVS-M Video coding fast sub-picture element movement estimating method claimed in claim 1, it is characterized in that the sub-pix fine search in Step5 is: in 1/2 pixel precision search, take the whole pixel motion vector of optimum as search center, 1/2 pixel is unit, do a little diamond search of step, preserve optimum 1/2 pixel and suboptimum 1/2 pixel position and matching error, 1/2 pixel of testing according to other needs of determining positions of optimum and suboptimum pixel; If optimum 1/2 pixel is adjacent with suboptimum 1/2 pixel and be not the search starting point, so only need increase again 1/2 pixel between consecutive points; If have one to be the search starting point in optimum or suboptimum l/2 pixel, need so to increase by two adjacent 1/2 pixels of test optimum point; If optimum and suboptimum 1/2 pixel is symmetrical and be not the search starting point, need to be tested remaining all 1/2 pixels; After complete 1/2 pixel motion vector of above-mentioned decision search, with the minimum match error o'clock search center as 1/4 pixel precision, 1/4 pixel is unit, does a little diamond search of step.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN2013100190213A CN103384324A (en) | 2013-01-19 | 2013-01-19 | Quick sub pixel motion estimation method for AVS-M video coding |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN2013100190213A CN103384324A (en) | 2013-01-19 | 2013-01-19 | Quick sub pixel motion estimation method for AVS-M video coding |
Publications (1)
Publication Number | Publication Date |
---|---|
CN103384324A true CN103384324A (en) | 2013-11-06 |
Family
ID=49492011
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN2013100190213A Pending CN103384324A (en) | 2013-01-19 | 2013-01-19 | Quick sub pixel motion estimation method for AVS-M video coding |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN103384324A (en) |
Cited By (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103327340A (en) * | 2013-06-20 | 2013-09-25 | 华为技术有限公司 | Method and device for searching integer |
CN105516726A (en) * | 2015-11-27 | 2016-04-20 | 传线网络科技(上海)有限公司 | Motion compensation matching method and system of video coding |
CN108271028A (en) * | 2016-12-30 | 2018-07-10 | 北京优朋普乐科技有限公司 | One sub-pixel all direction search method and device |
CN109660799A (en) * | 2017-10-12 | 2019-04-19 | 北京金山云网络技术有限公司 | Method for estimating, device, electronic equipment and storage medium in Video coding |
CN110753227A (en) * | 2019-10-23 | 2020-02-04 | 李思恒 | Video coding method based on interframe prediction |
CN112866719A (en) * | 2019-11-27 | 2021-05-28 | 北京博雅慧视智能技术研究院有限公司 | Rapid pixel-division prediction method for AVS2 |
Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN1741618A (en) * | 2004-09-03 | 2006-03-01 | 北京航空航天大学 | A fast sub-picture element movement estimating method |
CN102164283A (en) * | 2011-05-30 | 2011-08-24 | 江苏大学 | Sub-pixel motion estimation method based on AVS (audio video standard) |
-
2013
- 2013-01-19 CN CN2013100190213A patent/CN103384324A/en active Pending
Patent Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN1741618A (en) * | 2004-09-03 | 2006-03-01 | 北京航空航天大学 | A fast sub-picture element movement estimating method |
CN102164283A (en) * | 2011-05-30 | 2011-08-24 | 江苏大学 | Sub-pixel motion estimation method based on AVS (audio video standard) |
Non-Patent Citations (3)
Title |
---|
周城等: "《一种基于AVS的亚像素快速运动估计算法》", 《中国图象图形学报》, vol. 12, no. 9, 30 September 2007 (2007-09-30), pages 1520 - 1524 * |
宋雪桦等: "《基于AVS的快速亚像素运动估计算法》", 《计算机工程与设计》, vol. 33, no. 7, 31 July 2012 (2012-07-31), pages 2716 - 2720 * |
张新安等: "《基于AVS-M 的亚像素运动估计快速算法》", 《计算机工程》, vol. 36, no. 24, 31 December 2010 (2010-12-31), pages 214 - 216 * |
Cited By (10)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103327340A (en) * | 2013-06-20 | 2013-09-25 | 华为技术有限公司 | Method and device for searching integer |
CN103327340B (en) * | 2013-06-20 | 2016-06-15 | 华为技术有限公司 | A kind of integer searches method and device |
CN105516726A (en) * | 2015-11-27 | 2016-04-20 | 传线网络科技(上海)有限公司 | Motion compensation matching method and system of video coding |
CN105516726B (en) * | 2015-11-27 | 2019-04-09 | 传线网络科技(上海)有限公司 | The motion compensation matching process and system of Video coding |
US10448036B2 (en) | 2015-11-27 | 2019-10-15 | Chuanxian Network Technology (Shanghai) Co., Ltd | Motion compensation matching for video coding |
CN108271028A (en) * | 2016-12-30 | 2018-07-10 | 北京优朋普乐科技有限公司 | One sub-pixel all direction search method and device |
CN109660799A (en) * | 2017-10-12 | 2019-04-19 | 北京金山云网络技术有限公司 | Method for estimating, device, electronic equipment and storage medium in Video coding |
CN110753227A (en) * | 2019-10-23 | 2020-02-04 | 李思恒 | Video coding method based on interframe prediction |
CN112866719A (en) * | 2019-11-27 | 2021-05-28 | 北京博雅慧视智能技术研究院有限公司 | Rapid pixel-division prediction method for AVS2 |
CN112866719B (en) * | 2019-11-27 | 2022-09-23 | 北京博雅慧视智能技术研究院有限公司 | Rapid pixel-division prediction method for AVS2 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN103384324A (en) | Quick sub pixel motion estimation method for AVS-M video coding | |
CN101600108B (en) | Joint estimation method for movement and parallax error in multi-view video coding | |
CN101699865B (en) | Method for fast searching mass movement self-adapting sub pixel | |
CN102075760B (en) | Quick movement estimation method and device | |
CN101394566B (en) | Cross rhombic motion estimation searching method | |
CN101420617B (en) | Motion estimation searching method for cross hexagon | |
CN106604035B (en) | A method of the estimation for Video coding and compression | |
CN106507106B (en) | Video interprediction encoding method based on reference plate | |
CN102158709A (en) | Derivable motion compensation prediction method of decoding terminal | |
US20230042575A1 (en) | Methods and systems for estimating motion in multimedia pictures | |
CN112235570B (en) | Depth prediction method based on precoding and intra-frame direction prediction method | |
CN102088605B (en) | Rapid interlayer prediction selection method for scalable video coding | |
CN103384332A (en) | Improved algorithm for HFPS algorithm of AVS video coding | |
Qaralleh et al. | Fast variable block size motion estimation by adaptive early termination | |
CN103260020A (en) | Quick integer pixel motion estimation method of AVS-M video coding | |
CN106658024B (en) | A kind of quick method for video coding | |
Podder et al. | Efficient coding strategy for HEVC performance improvement by exploiting motion features | |
CN105611299B (en) | A kind of method for estimating based on HEVC | |
CN103096093A (en) | Error concealment method of H. 246 video P frame | |
Xiang et al. | Fast intra mode decision for HEVC | |
CN103402084A (en) | AVS (Audio Video coding Standard)-based motion estimation fast algorithm | |
Zheng et al. | A new multihypothesis prediction scheme for compressed video sensing reconstruction | |
CN102263958A (en) | method and device for obtaining initial point based on H264 motion estimation algorithm | |
Khairy et al. | Low complexity for scalable video coding extension of H. 264 based on the complexity of video | |
CN103475883A (en) | HEVC motion estimation early termination method based on division of motion zones |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
C02 | Deemed withdrawal of patent application after publication (patent law 2001) | ||
WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20131106 |