CN103402084A - AVS (Audio Video coding Standard)-based motion estimation fast algorithm - Google Patents
AVS (Audio Video coding Standard)-based motion estimation fast algorithm Download PDFInfo
- Publication number
- CN103402084A CN103402084A CN201310306762XA CN201310306762A CN103402084A CN 103402084 A CN103402084 A CN 103402084A CN 201310306762X A CN201310306762X A CN 201310306762XA CN 201310306762 A CN201310306762 A CN 201310306762A CN 103402084 A CN103402084 A CN 103402084A
- Authority
- CN
- China
- Prior art keywords
- search
- point
- algorithm
- sad
- sds
- 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
Landscapes
- Compression Or Coding Systems Of Tv Signals (AREA)
Abstract
The invention provides an AVS (Audio Video coding Standard)-based motion estimation fast algorithm. According to the fast algorithm, a successive elimination algorithm-based square-diamond search (SEA-SDS) strategy is designed in combination with the SEA based on the SDS. In order to fully utilize the temporal spatial correlation of motion vector distribution and improve the search performance, a predictive adaptive square-diamond search algorithm using successive elimination (PA-SEA-SDS) is provided in combination with prediction of a search start point, adaptive determination of a search mode and selection of a static block, and the steps of the algorithm are as shown in the attached drawings. The AVS-based motion estimation fast algorithm has the positive effects that the algorithm is superior to the conventional fast motion estimation algorithms such as DS (Diamond Search) and SDS in both search accuracy and search speed, an estimation effect which is equivalent to that of FS (Free Search) is achieved with low cost and the algorithm has extremely strong adaptability and robustness.
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 motion-estimation algorithm based on AVS.
Background technology
The digital audio/video industry is one of information industry three large parts, the digital audio/video industry size of China is huge, but intellectual property is weak, and this field, for a long time by MPEG-2, MPEG-4 with H.264 wait foreign technology standard monopolization, adopts these standards all to need to pay high patent fee.For the patent fee burden that reduces domestic audio frequency and video industry, the core competitiveness that promotes domestic enterprise, the Ministry of Information industry is in June, 2002 approval establishment " digital audio/video coding techniques standard operation group ", be engaged in scientific research institution and the enterprise of the research and development of digital audio/video coding techniques in the United Nations, demand for domestic audio frequency and video industry, proposed the autonomous information source coding standard of China-" information technology advanced audio/video coding " series standard, be called for short AVS(Audio Video coding Standard).
AVS with H.264 compare, have 4 large advantages: (1) performance is high, code efficiency is 2.4 times of MPEG-2, with H.264 suitable; (2) complexity is low, encoder complexity be equivalent to H.264 70%, decoding complex degree be equivalent to H.264 30%; (3) software and hardware realizes that cost is all lower than H.264; (4) patent fee is low, only collects the software and hardware patent fee of 1 yuan.The use of AVS, not only can save the patent fee of tens dollars every year, can also drive the chip industry development of China, accelerates the problem of solution China IT enterprises coreless technology.
In the AVS Video coding, estimation (Motion Estimation, ME) is the module of amount of calculation maximum, accounts for 76% of whole encoding calculation amount.In the motion estimation algorithm of AVS, to every kind of block mode, all to carry out respectively the search of whole pixel, half-pix and 1/4 pixel.Whole pixel motion estimates to account for the amount of calculation of whole coding 68%, how to improve the efficiency of estimation, make the search procedure of motion estimation algorithm quicker, more efficient be a focus of Video coding research field always.
The fast search algorithm of estimation has a variety of at present, typical algorithm has: three-step approach (Three Step Search, TSS), new three-step approach (Novel Three Step Search, NTSS), four step rule (Four Step Search, FSS), diamond search method (Diamond Search, DS), hexagon search method (Hexagon-based Search, HBS), second order logarithm method (2-D LOGS) and being combined with between them etc.First three methods easily is absorbed in local optimum, and search speed is fast not.The DS algorithm is first with large rhombus, to mate, and carries out the coarse positioning of Best Point, and search procedure can not be absorbed in local minimum, after coarse positioning finishes, re-uses little rhombus and accurately locates.The HBS algorithm is similar to DS, and just search pattern is different, and it has used than DS algorithm future position still less, has therefore improved the search speed of algorithm.
Summary of the invention
Need more efficient rapid motion estimating method for AVS, the deficiency that exists in order to solve existing motion estimation algorithm, the present invention discloses a kind of fast motion-estimation algorithm based on AVS.
The present invention is at square-rhombus algorithm (Square-Diamond Search, SDS) on basis,, in conjunction with successive elimination algorithm (Successive Elimination Algorithm, SEA), designed a kind of square-rhombus based on successive elimination (search strategy of SEA-SDS).In order to take full advantage of the temporal correlation of motion vector distribution, improve search performance, therefore, on the basis of SEA-SDS, determine the selection of search pattern and static block in conjunction with search starting point prediction, self adaptation, a kind of adaptive square-diamond search algorithm based on successive elimination (A Predictive Adaptiv Square-Diamond Search Algorithm using Successive Elimination, PA-SEA-SDS) of measurable search starting point have been proposed.
Algorithm steps of the present invention is as follows:
The Stepl static block detects: Zerohunt vector position (0,0) point, calculate this point
SADValue
SAD 0If,
SAD 0＜
T 0, judge that this piece is static block.(0,0) position is the optimal motion vector, changes Step6 over to; Otherwise carry out Step2;
The Step2 type of sports detects: also have the sub-block of same position in reference frame to draw the predictive vector range L of current block according to top, left, upper right side sub-block, if meet L≤L
1, illustrate that current block is little moving mass, changes Step4 over to; If meet L
l＜L≤L
2, current block is the middle motion piece, changes Step5 over to; As L＞L
2The time, current block belongs to large moving mass, need to carry out the starting point prediction, changes Step3 over to;
Step3 starting point prediction:, in the motion vector of the time and space of current block 4 the adjacent piece Select Error minimums starting point as the current block estimation, change Step4 over to;
Step4 SDP search: centered by current MBD point, use SDP to mate., if the MBD point is positioned at center, illustrate that current point is the optimal motion vector, changes Step6 over to; Otherwise make this MBD point be starting point, repeat Step4, continue to use SDP to mate;
Sep5 SAE-SDS search:, take current point as the search initial point, use SEA-SDS search method to search for, obtain the optimal motion vector, change Step6 over to;
The Step6 algorithm finishes.
Described MBD point is smallest blocks distortion (Minimum Block Distortion, MBD) point, i.e. optimal match point.Its matching criterior is the actual absolute error of using and (Sum of Absolute Difference, SAD) in estimation, is defined as follows:
In formula, (
i,
j) the expression displacement vector,
f kWith
f k-1The gray value that represents respectively present frame and reference frame, M * N are the size of macro block,
SAD(
i,
j) minimum point is exactly optimal match point.
Compared with the prior art, advantage of the present invention and good effect are: PA-SEA-SDS algorithm, all surpassed fast motion estimation algorithm such as DS, SDS in the past on search precision or search speed, with less cost, obtain the estimation effect suitable with FS, had very strong adaptivity and robustness.
Below from the judgement of the static block that affects motion estimation performance, the judgement of type of sports, prediction and four parts of search strategy of search starting point, the present invention is elaborated.
The judgement of static block.
In general, the motion vector of video sequence has center-biased property.The common high concentration of the motion vector of sequence of video images is distributed near the center of search window, and wherein static block occupies very large proportion, and zero vector is also very favorable for coding.Design threshold for this reason
T, at first detect (0,0) vector before search.When weighing the piece distortion factor, algorithm of the present invention uses
SADAs the criterion of weighing error in judgement.
If meet
SAD(0,0)＜
T, think that this piece belongs to static block, its optimal motion vector is exactly (0,0), directly ends search this moment.Experimental results show that and work as
T=512 o'clock, probability of miscarriage of justice was minimum.Be that zero vector point can be completed search by the method for static block, can greatly save amount of calculation.But due to the complexity of motion sequence, obviously fixing threshold value does not have universality,, according to the correlation of motion, uses new static block decision method for this reason.
According to time and the spatial coherence of macro block, the search starting point that can utilize the motion vector of adjacent block to predict current block.In general, the image block of the adjacent block of the own coding of locus and former frame can be used to predict the motion vector of current block.Experiment shows, top in current block and frame, left, top-right in addition in reference frame the correlation of the piece identical with the current block position the strongest, with the correlation of other positions a little less than.
In algorithm of the present invention, adopted with top in this piece frame with the left piece
SADIntermediate value as judging whether macro block is the new threshold value of static block
T 0, namely
T 0=
Median(
SAD 1,
SAD 2).Wherein
SAD 1,
SAD 2Be respectively current block left side piece and upper block
SADValue.Because the threshold value that piece is corresponding is adaptive change, so this threshold value setting mode is stronger than the mode adaptability that adopts fixed threshold, the judgement static block is more accurate.
The judgement of type of sports.
Due to the globality of moving object and the continuity of video motion, the motion vector of adjacent motion vector and frame corresponding position, front and back must have the correlation on time and space.If utilize this temporal correlation of motion vector just can make prediction to the motion conditions of current block, be beneficial to and carry out subsequently adaptable search.Experiment shows, top in current block and frame, left, top-right in addition in reference frame the correlation of the piece of same position the strongest, these 4 pieces of algorithm picks of the present invention are predicted as candidate point.Reference block is (namely the piece in figure 1,2,3,4) as shown in Figure 1.
Definition motion vector set V={V
0, V
1, V
2, V
3, V
4, V
0=(0,0), V
i=(
x i ,
y i ),
i=1,2,3,4 represent respectively the motion vector of the adjacent piece of 4 space-times, the absolute value distance of definition motion vector
l i For:
Wherein,
x i It is motion vector
xDurection component,
y i It is motion vector
yDurection component.
All candidate vectors in the motion vector set, make L=max{
l 0,
l 1,
l 2,
l 3,
l 4, L can reflect the size of macro block motion amplitude.Set type of sports discrimination threshold L
1And L
2So that type of sports is classified, and L
l≤ L
2:
(1) as L≤L
1The time, judge that current block is little moving mass, only need to be the density search of little step-length near search starting point;
(2) when meeting L
l＜L≤L
2, current block is the middle motion piece, needs to do the search of large step-length near search starting point;
(3) as L＞L
2The time, current block is large moving mass, illustrates that the motion amplitude of this image block may be larger, need to do the search of little step-length near the search starting point after prediction.
After having judged the macro block type of sports,, targetedly motion estimation algorithm suitable according to drawn motion conditions choice for use.In the Video coding of reality, can set flexibly type decision threshold value L according to the type of image sequence and the needs of encoder-side
lAnd L
2Value.By a large amount of experiments, the present invention is set as L with the decision threshold of type of sports
l=1, L
2=3 differentiate the macro block of 3 kinds of type of sports.
The prediction of search starting point.
Initial search point is set according to the type of sports of current block.When if current block is little moving mass or middle motion piece, the optimal motion vector is positioned near zone less (0,0), need not carry out the starting point prediction; If current block is large type of sports, it is larger that its optimal motion vector often departs from central point, and the initial search point after this moment prediction is more near the true motion vector of object.
Concrete grammar of the present invention is: the motion vector of establishing current block and 4 space-time adjacent blocks is respectively V
0, V
1, V
2, V
3, V
4Calculate respectively the piece distortion factor (BDM) corresponding to piece of these 5 motion vector indications, the index of tolerance BDM size adopt absolute error and
SADThe piece distortion factor that calculates is made as respectively
SAD 0,
SAD 1,
SAD 2,
SAD 3,
SAD 4, calculate the motion vectors V of current block search starting point as shown in (3) formula:
Search strategy.
In order to search fast and accurately optimum vector to various types of, the present invention has adopted two kinds of templates: large square template (LSP template) and little cross template (SDP search pattern), as shown in Figure 2.
The off-centring of search starting point
In general, the motion vector distribution of the actual video sequence image that obtains has certain rule, namely motion vector usually all high concentration zero vector and near, be called off-centring, namely most of motion vector be zero vector or the motion very little.The distribution of motion vector is except very intensive near central point, and the motion vector distribution on level and vertical direction is more intensive than other direction.
The off-centring characteristic plays very important effect in searching algorithm, the distribution character that the model of searching algorithm must meet motion vector just may reach higher search speed, otherwise not only can cause large operand, but also likely cause algorithm not restrained.
Successive elimination (SEA) algorithm SEA(Successive Elimination Algorithm) be that W.Li and E.Safari proposed in nineteen ninety-five.This algorithm is outstanding harmless fast algorithm, on the basis of not losing any performance, can effectively accelerate the speed of searching for.SEA utilize a succinct inequality reduce point to be matched absolute error and
SADAmount of calculation, avoid the hopeless position that becomes optimal match point is done further to calculate.SEA think certain a bit (
i,
j) precondition that likely becomes optimal match point is:
Wherein, △
Sum0 is the absolute value sum of block to be encoded each point gray scale; Min
SADFor minimum in the point of having searched for
SAD△
Sum(
i,
j) be to be matched (
i,
j) the absolute value sum of each point gray scale.
Premature termination a bit (
i,
j) coupling computational process condition be whether (4) formula is set up.While namely searching for new match point, first judge whether to meet (4) formula at every turn,, if meet, calculate
SAD(
i,
j), directly take off a bit otherwise just skip this point.If
SAD(
i,
j) than current Min
SADLittle, upgrade Min
SADAnd respective vectors: Min
SAD=
SAD(
i,
j).Last Search Results is exactly Min
SADWith its corresponding motion vector.
The successive elimination algorithm flow process as shown in Figure 3.
SEA-SDS search strategy is based on the offset characteristic of SEA algorithm above-mentioned and motion vector, adopt large square (LSP) shown in Figure 2 and two kinds of search patterns of crosslet (SDP), the present invention has designed a kind of new motion estimation search strategy-SEA-SDS searching algorithm.As shown in Figure 4, the concrete steps of algorithm are as follows for SEA-SDS searching algorithm flow process:
Step1, take the current block center as initial search point, uses the LSP template.First calculate LSP region of search central point
SADValue, then utilize the SEA algorithm to mate calculating to 4 points on every side, if 4 points all do not meet Rule of judgment (4) formula of SEA algorithm or have at least any to meet this condition on every side, and MBD(Minimum Block Distortion) point be positioned at central point, forward Step3 to; Otherwise carry out Step2;
The MBD point that the previous step of using Step2 finds, as central point, calculates with new LSP and SEA, if 3 points all do not meet (4) formula on every side, perhaps the MBD point is positioned at central point, carries out Step3; Otherwise repeat Step2;
The MBD point that the previous step of using Step3 finds, as central point, is changed to SDP with LSP, utilizes SEA to calculate at 5 some places, if 4 points all can not meet (4) formula on every side, the central point position of SDP is corresponding optimum movement vector, and search finishes; Otherwise find out the MBD point, turn Step4;
Step4 centered by the minimum MBD point that previous step finds, continue to use SDP template and SEA search (note the point of search for need not search for) again, finds out new MBD point in these points, and this position namely corresponds to optimum movement vector, searches for end.
Description of drawings
Fig. 1 is the reference block schematic diagram that 4 pieces of algorithm picks of the present invention are predicted as candidate point.
Fig. 2 is large square template (LSP) and little cross template (SDP) schematic diagram that the present invention adopts.
Fig. 3 is SEA successive elimination algorithm flow chart.
Fig. 4 is SEA-SDS algorithm flow chart.
Fig. 5 is PA-SEA of the present invention-SDS algorithm flow chart.
Embodiment
PA-SEA-SDS arthmetic statement, owing to having adopted square-rhombus template, combines successive elimination algorithm, and basic SEA-SDS algorithm has had certain raising than the SDS algorithm on search speed.But due to the temporal correlation that does not utilize motion vector distribution, its performance is difficult people's will to the greatest extent still.Improvement in performance is carried out to rudimentary algorithm in the aspects such as the present invention predicts from search starting point, self adaptation is determined search pattern, a kind of adaptive square-diamond search algorithm based on successive elimination (A Predictive Adaptive Square-Diamond Search Algorithm using Successive Elimination, PA-SEA-SDS) of measurable search starting point are proposed.
PA-SEA-SDS algorithm idea adopts a kind of search pattern from coarse to fine, first by the threshold value of setting, piece is divided different type of sports, then for the different search pattern of different type selecting.
PA-SEA-SDS algorithm flow algorithm flow as shown in Figure 5.
PA-SEA-SDS algorithm steps algorithm steps is as follows:
The Stepl static block detects: Zerohunt vector position (0,0) point, calculate this point
SADValue
SAD 0If,
SAD 0＜
T 0, judge that this piece is static block.(0,0) position is the optimal motion vector, changes Step6 over to; Otherwise carry out Step2;
The Step2 type of sports detects: also have the sub-block of same position in reference frame to draw the predictive vector range L of current block according to top, left, upper right side sub-block, if meet L≤L
1, illustrate that current block is little moving mass, changes Step4 over to; If meet L
l＜L≤L
2, current block is the middle motion piece, changes Step5 over to; As L＞L
2The time, current block belongs to large moving mass, need to carry out the starting point prediction, changes Step3 over to;
Step3 starting point prediction:, in the motion vector of the time and space of current block 4 the adjacent piece Select Error minimums starting point as the current block estimation, change Step4 over to;
Step4 SDP search: centered by current MBD point, use SDP to mate., if the MBD point is positioned at center, illustrate that current point is the optimal motion vector, changes Step6 over to; Otherwise make this MBD point be starting point, repeat Step4, continue to use SDP to mate;
Sep5 SAE-SDS search:, take current point as the search initial point, use SEA-SDS search method to search for, obtain the optimal motion vector, change Step6 over to;
The Step6 algorithm finishes.
Algorithm Analysis.
The characteristics of PA-SEA-SDS algorithm maximum are exactly after macroblock partitions is different type of sports, the search strategy of all kinds of macro blocks of reasonable arrangement, and as long as with require to coincide, can finish to search for.Thereby this way of search not only can have very high matching precision, can find rapidly the Optimum Matching macro block simultaneously, meets the real-time requirement of data.
Experimental result and analysis.
Estimate the quality of a motion estimation algorithm, mainly see the effect of its coupling and the time complexity of search.Generally matching effect is that average peak signal to noise ratio PSNR by image weighs.Time complexity can compare by search point and search time.Due to the impact that is subject to operation platform and other factors search time.At present modal is that comparison search is counted, and namely carries out matching ratio number of times in search procedure.
In order to verify PA-SEA proposed by the invention-SDS algorithm validity, the present invention has used several more representative rapid movement algorithms: FS, TSS, and DS and SDS, compare experiment with PA-SEA-SDS algorithm under identical condition.The sub-block size of using is 16 * l6, and search window size is ± 15 pixels, and matching criterior is used
SAD
Choose the image sequence of 4 representative CIF forms in experiment: Calire, Football, Tennis, Mobile.Calire is typical head shoulder sequence, and image motion is small mild; Football belongs to large motion sequence; Comprise rapid movement and the scene switching of object in the Tennis sequence; The translation that had both contained camera lens in the Mobile sequence also comprises the movement of object, and background is complicated.Front 100 frames in above-mentioned 4 image sequences, carry out estimation every frame, calculates the mean value of its PSNR and search point.The mean value of the search point that obtains of experiment and the mean value result of PSNR are respectively in Table 1 and table 2.
The table 1 average search comparison (unit: inferior/pixel) of counting
The average search that table 1 has been listed each algorithm is counted and the comparison of each algorithm and FS search speed.Observation table 1 can find, for the Calire sequence, FS need to search for 225 points, and DS need to search for 13.12 points, and the SDS algorithm also needs 2.65 points, and PA-SEA-SDS on average only need search for 1.92 points, can find optimum vector.Also effectively reduced search point for large motion sequence PA-SEA-SDS, be compared to SDS, PA-SEA-SDS, reduced by 2.67 search points in the Football sequence, reduced by 2.2 search points in the Tennis sequence.
As can be seen from Table 1, algorithm of the present invention, no matter for the video image of which kind of type of sports, can both obviously reduce the average search of finding optimum vector and count, and is the fastest algorithm of search speed.
Table 2 average PSNR value is (unit: dB)
Table 2 has been listed the PSNR mean value of the reconstructed image of each algorithm, and the comparison of each algorithm and FS precision.Therefrom can find out, the mean P SNR value of FS is the highest, and the search precision that FS is described is the highest.The search precision of PA-SEA-SDS in all sequences all, higher than other fast algorithm, has a little decline than FS algorithm, but on almost not impact of picture quality.This explanation PA-SEA-SDS algorithm under the prerequisite that guarantees search quality, has improved the search speed of estimation.
Table 1 and table 1 reflect: no matter PA-SEA-SDS algorithm is for image change mild video such as Calire sequence, still change violent video such as Football sequence, can keep good performance aspect search speed and PSNR two, have very strong adaptability and robustness.
Should be understood that; above-mentioned explanation is not limitation of the present invention, and the present invention also is not limited in above-mentioned giving an example, those skilled in the art make in essential scope of the present invention modification; distortion, interpolation or replacement, also should belong to protection scope of the present invention.
Claims (7)
1. based on the fast motion-estimation algorithm of AVS, its feature comprises the following steps:
The Stepl static block detects: Zerohunt vector position (0,0) point, calculate this point
SADValue
SAD 0If,
SAD 0＜
T 0, judge that this piece is static block, (0,0) position is the optimal motion vector, changes Step6 over to; Otherwise carry out Step2;
The Step2 type of sports detects: also have the sub-block of same position in reference frame to draw the predictive vector range L of current block according to top, left, upper right side sub-block, if meet L≤L
1, illustrate that current block is little moving mass, changes Step4 over to; If meet L
l＜L≤L
2, current block is the middle motion piece, changes Step5 over to; As L＞L
2The time, current block belongs to large moving mass, need to carry out the starting point prediction, changes Step3 over to;
Step3 starting point prediction:, in the motion vector of the time and space of current block 4 the adjacent piece Select Error minimums starting point as the current block estimation, change Step4 over to;
Step4 SDP search: centered by current MBD point, use SDP to mate,, if the MBD point is positioned at center, illustrate that current point is the optimal motion vector, changes Step6 over to; Otherwise make this MBD point be starting point, repeat Step4, continue to use SDP to mate;
Sep5 SAE-SDS search:, take current point as the search initial point, use SEA-SDS search method to search for, obtain the optimal motion vector, change Step6 over to;
The Step6 algorithm finishes.
2. the fast motion-estimation algorithm based on AVS according to claim 1, is characterized in that described in Stepl
SAD, for the actual absolute error of using and (Sum of Absolute Difference, SAD) in estimation, be defined as follows:
In formula, (
i,
j) the expression displacement vector,
f kWith
f k-1The gray value that represents respectively present frame and reference frame, M * N are the size of macro block.
3. the fast motion-estimation algorithm based on AVS according to claim 1, is characterized in that described in Stepl
T 0For the decision threshold of static block,
T 0=
Median(
SAD 1,
SAD 2), wherein
SAD 1,
SAD 2Be respectively current block left side piece and upper block
SADValue.
4., at the fast motion-estimation algorithm based on AVS claimed in claim 1, it is characterized in that the L described in Step2 is the type of sports discrimination threshold, is set as L
l=1, L
2=3, L=max{
l 0,
l 1,
l 2,
l 3,
l 4,
l i Absolute value distance for motion vector:
Wherein,
x i It is motion vector
xDurection component,
y i It is motion vector
yDurection component.
5. at the fast motion-estimation algorithm based on AVS claimed in claim 1, the time that it is characterized in that the current block described in Step3 and the adjacent piece in 4, space be current block top, left, top-right with reference frame in the piece identical with the current block position.
6., at the fast motion-estimation algorithm based on AVS claimed in claim 1, it is characterized in that the MBD point described in Step4 is smallest blocks distortion (Minimum Block Distortion, MBD) point, i.e. optimal match point.
7. at the fast motion-estimation algorithm based on AVS claimed in claim 1, it is characterized in that the SEA described in Step5-SDS search method is successive elimination (Successive Elimination Algorithm, square-rhombus SEA) (Square-Diamond Search, SDS) search method (SEA-SDS).
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201310306762XA CN103402084A (en) | 2013-07-22 | 2013-07-22 | AVS (Audio Video coding Standard)-based motion estimation fast algorithm |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201310306762XA CN103402084A (en) | 2013-07-22 | 2013-07-22 | AVS (Audio Video coding Standard)-based motion estimation fast algorithm |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN103402084A true CN103402084A (en) | 2013-11-20 |
Family
ID=49565599
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201310306762XA Pending CN103402084A (en) | 2013-07-22 | 2013-07-22 | AVS (Audio Video coding Standard)-based motion estimation fast algorithm |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN103402084A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106878756A (en) * | 2017-03-08 | 2017-06-20 | 苏州达尔普工业控制有限公司 | Robot motion estimates and backoff algorithm |
| CN110493602A (en) * | 2019-08-19 | 2019-11-22 | 张紫薇 | Video coding fast motion searching method and system |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20050201627A1 (en) * | 2004-03-11 | 2005-09-15 | Yi Liang | Methods and apparatus for performing fast mode decisions in video codecs |
| CN101227611A (en) * | 2008-01-31 | 2008-07-23 | 上海广电（集团）有限公司中央研究院 | AVS-based motion estimation apparatus and searching method |
| CN101699865A (en) * | 2009-10-28 | 2010-04-28 | 山东大学 | Method for fast searching mass movement self-adapting sub pixel |
-
2013
- 2013-07-22 CN CN201310306762XA patent/CN103402084A/en active Pending
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20050201627A1 (en) * | 2004-03-11 | 2005-09-15 | Yi Liang | Methods and apparatus for performing fast mode decisions in video codecs |
| CN101227611A (en) * | 2008-01-31 | 2008-07-23 | 上海广电（集团）有限公司中央研究院 | AVS-based motion estimation apparatus and searching method |
| CN101699865A (en) * | 2009-10-28 | 2010-04-28 | 山东大学 | Method for fast searching mass movement self-adapting sub pixel |
Non-Patent Citations (1)
| Title |
|---|
| 焦蓬蓬: "AVS视频编码中整数变换与运动估计研究", 《中国优秀硕士学术论文电子期刊网》 * |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106878756A (en) * | 2017-03-08 | 2017-06-20 | 苏州达尔普工业控制有限公司 | Robot motion estimates and backoff algorithm |
| CN110493602A (en) * | 2019-08-19 | 2019-11-22 | 张紫薇 | Video coding fast motion searching method and system |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN112584177B (en) | Coding and decoding method and equipment thereof | |
| RU2381630C2 (en) | Method and device for determining block conformity quality | |
| US10051282B2 (en) | Method for obtaining motion information with motion vector differences | |
| CN101394566B (en) | Cross rhombic motion estimation searching method | |
| CN101420617B (en) | Motion estimation searching method for cross hexagon | |
| JP2005535228A (en) | Method and apparatus for performing high quality fast predictive motion search | |
| CN102075760A (en) | Quick movement estimation method and device | |
| CN103384327A (en) | AVS fast mode selection algorithm based on adaptive threshold | |
| CN103384324A (en) | Quick sub pixel motion estimation method for AVS-M video coding | |
| CN101237580A (en) | Integer pixel quick mixing search method based on center prediction | |
| CN103402084A (en) | AVS (Audio Video coding Standard)-based motion estimation fast algorithm | |
| CN101309424B (en) | Quick movement estimating method | |
| WO2013013548A1 (en) | Initial point acquisition method and device based on h.264 motion estimation algorithm | |
| JPH10304371A (en) | Moving vector detecting device | |
| CN101094404A (en) | Motion estimation algorithm of self-adapting stretching template | |
| Ji et al. | Reduced computation using adaptive search window size for H. 264 multi-frame motion estimation | |
| Wang et al. | A novel hexagon diamond search algorithm for motion estimation | |
| He et al. | Improved hexagon-based searching algorithm for fast motion estimation | |
| CN103384330A (en) | Quick AVS motion estimation method | |
| Wang et al. | Optimization of UMHexagonS algorithm for AVS+ | |
| Zu-Jue et al. | A fast motion estimation optimization for H. 264/AVC encoder | |
| Cai et al. | Video Coding Algorithm for Block Motion Estimation | |
| CN112866691A (en) | Inter-frame fast mode decision method for video coding | |
| Wang et al. | An improved motion-search method based on pattern classification | |
| JP2005328115A (en) | Motion vector detection apparatus and motion vector detection program |
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: 20131120 |