CN1181691C - Vidio motion estimation method - Google Patents
Vidio motion estimation method Download PDFInfo
- Publication number
- CN1181691C CN1181691C CNB031151337A CN03115133A CN1181691C CN 1181691 C CN1181691 C CN 1181691C CN B031151337 A CNB031151337 A CN B031151337A CN 03115133 A CN03115133 A CN 03115133A CN 1181691 C CN1181691 C CN 1181691C
- Authority
- CN
- China
- Prior art keywords
- vector
- block
- piece
- motion
- motion vector
- 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.)
- Expired - Lifetime
Links
Images
Landscapes
- Compression Or Coding Systems Of Tv Signals (AREA)
Abstract
The present invention relates to a method for carrying out motion estimation on the basis of the space coherence of a motion vector field in a searching mode. The present invention comprises the steps that a group of candidate vectors are determined on the basis of a block matching and searching order, according to the motion vectors of some nearby blocks and by an updating mechanism; for each candidate vector, the errors of the current pixel block and a matched reference pixel block thereof are worked out; the errors of the current pixel block and a pixel block at the same position in a reference frame are worked out; the vector with the minimal error value is determined as a motion vector; the estimated motion vector is treated by median filtering to smoothen a vector field. The present invention has the advantages of small calculation amount and low cost. Because the correctness of motion estimation is enhanced, the true motion of an object can be well described by the motion vector field.
Description
Technical field
The invention belongs to the Digital Video Processing technical field, the method for estimation is carried out in particularly a kind of utilization based on the temporal correlation of the motion vector field of piece match search order.
Background technology
All the time, motion estimation techniques is widely used in fields such as video compression coding and video format conversion.In order to reduce the needed bandwidth of digital of digital video data that transmits given picture quality as far as possible, multiple different video compression algorithm has been developed and has been used for compressed video data.The several multimedia technology standard committee set up and proposed to be used to encode/standard of compressed audio and video data.In the middle of this, be widely known by the people and widely accepted international standard provide by Motion Picture Experts Group (MPEG), comprise MPEG-1 and Moving Picture Experts Group-2 etc.
Method for video coding mainly comprises compression and interframe compression in the frame, and wherein the interframe compression is the main source of its compression ratio, so existing video compression standard (MPEG1/2/4, H.261/3 or the like.) most interframe compression scheme that adopts based on estimation.Its principle is exactly earlier present frame to be divided into the identical piece of some sizes, then to each piece (current block) the most similar with it match block of search in a certain size window of reference frame.The alternate position spike of current block and match block is called motion vector, the pixel difference is called residual block.Because the pixel near 0 in the residual block is a lot, by dct transform, quantification, entropy coding, just can increase substantially compression ratio.In said process, estimation is exactly the link of search best matching blocks.Obviously, estimation is not only the most time-consuming, and directly influences compression efficiency, is the key and the bottleneck of video compression.
The scan format conversion is the common issue with that calculating video, high definition TV and the traditional tv under the current network environment exists, and scan format need be changed between different systems.Deinterleaving is a basic demand of in the practice all scan formats being changed.Propose the algorithm of many deinterleavings, included the method for simple space deinterleaving and the method for senior motion compensation deinterleaving.The key technology of motion compensation deinterleaving is an estimation, and the complexity of estimation and performance will directly determine the complexity and the performance of motion compensation deinterleaving.The deinterleaving technology with technology of video compressing encoding to estimation require differently, it emphasizes that more motion vector meets the real motion of object.
The method of existing many estimation all exists some defectives: amount of calculation is excessive, and the cost of realization is big, such as full-search algorithm; Suppose the monotonicity of error curved surface, the as easy as rolling off a log local optimum that is absorbed in is such as some classical block matching methods, as three-step approach.The most important thing is that the method for these estimation all can not be described the real motion of object well.
The patent No. is the method that 01801368.6 patent of invention discloses a kind of estimation, this method is at first determined one group of candidate vector for the current pixel piece, select the starting point of best candidate vector then, search for to the end motion vector in presumptive area as search.The patent No. method of another kind of estimation that has been 01100544.0 patent disclosure, this method is at first determined the search starting point of an optimum for the current search piece, adopt diamond search then, and stops search procedure adaptively.These two kinds of methods all adopt fixing from left to right order from top to bottom to carry out the piece match search, utilize the correlation of the motion vector field under this order to reduce the amount of calculation of estimation and improve search speed and accuracy.The number and the position of the candidate vector of these two kinds of methods are all fixed, and the motion vector that estimates is no longer handled.Adopt permanent order to carry out the correlation that the piece match search can not make full use of the motion vector space, the performance of estimation is also unequal in the distribution of image space.
Summary of the invention
Purpose of the present invention is exactly defective or the deficiency that exists at prior art, and the temporal correlation of a kind of utilization based on the motion vector field of piece match search order is provided, and carries out the method for estimation.The motion vector that adopts this method to estimate can be described the real motion of object better.
This method may further comprise the steps: (a) based on the order of piece match search, according near the current block several motion vector and adopt a kind of update mechanism to determine one group of candidate vector; (b), calculate the error of the reference pixel block of current pixel piece and its coupling to each described candidate vector; Calculate the error of same position block of pixels in current pixel piece and the reference frame simultaneously, the vector of choosing above error amount minimum is as motion vector; (c) motion vector that estimates is carried out the level and smooth vector field of medium filtering.Certainly, represent in the step (b) that the motion vector of same position block of pixels in the current pixel piece match reference frame can be used as another candidate vector in the step (a), calculate the matching error of these candidate vector correspondences more simultaneously.
The order of piece match search can be from top to bottom from left to right, from top to bottom from left to right, from top to bottom from right to left, from top to bottom from right to left in the step (a); From top to bottom alternately from left to right and from right to left from top to bottom alternately from left to right and from right to left; From around to central authorities.
Step (a) comprises following steps: (1) according to the motion vector of two pieces of having handled of selective sequential of piece match search, one undressed motion vector, three vectors are as the preliminary candidate vector of determining altogether.Two treated pieces are selected adjacent with the current block level and vertical adjacent piece; One the undressed piece of selecting current block 45 degree diagonals.(2) according to a kind of update mechanism the preliminary candidate vector of determining is upgraded.
Update mechanism in the step (2) in the step (a) is exactly: at first to the motion vector of two in the preliminary candidate vector of the determining pieces of having handled according to the order of match search according to the selective sequential that replaces one of them.Then, in the vector space of a preset range, select a vector randomly, and this vector is added on selected that candidate vector of tentatively determining that upgrades, form the vector after upgrading.Other two that do not upgrade in vector after this upgrades and the preliminary candidate vector of determining constitute one group of candidate vector together.
Calculate error packet in the step (b) and contain following steps: (1) enlarges the block of pixels of pending present frame/field and mates the border of the block of pixels of corresponding reference frame by candidate vector, and it is carried out sub-sampling; (2), calculate the error of the reference pixel block of current pixel piece and its coupling to the piece pixel behind sub-sampling.
The error of calculation can not carried out sub-sampling again in the step (b), directly calculates the error of the reference pixel block of current pixel piece and its coupling.
A kind of simple realization that reduces amount of calculation of the error amount of the reference pixel block of calculating current pixel piece and its coupling comprises following steps in the step (b): (1) is to the pixel summation of current pixel piece respective number; (2) pixel of reference frame block of pixels respective number is sued for peace; (3) to above-mentioned two and ask absolute difference, just obtain motion estimation error.Certainly, also can directly calculate the error of the sad value of these pixels as estimation.
Step (c) comprises the steps: that (1) adopt the 2 dimension medium filterings of 3*3 respectively to the level of block motion vector and vertical component; (2) result to medium filtering output handles: the motion vector that the horizontal component of medium filtering output and vertical component constitute, if this vector is not any one that participates in 8 current block adjacent block motion vectors of medium filtering, the motion vector of getting original current block is as output, otherwise the result who gets medium filtering is as output.
This invention can be carried out estimation to the vision signal of various forms.Especially, this invention is applicable to digital hdtv signal is carried out estimation.
For most of images, the object of motion is bigger than piece usually, and the motion object have inertia, this just means that motion vector field has very big correlation in time with on the space.Utilize this correlation can find match block faster, more accurately, reduce amount of calculation greatly.The search order of general block matching algorithm is from top to bottom from left to right, this invention can be adopted many other search orders, and select candidate vector based on search order, can estimate motion more accurately, thereby make the motion vector that estimates more near the real motion of object.In the estimation, computation of match errors is the amount of calculation the best part, and this invention provides an advantages of simplicity and high efficiency implementation method, has reduced amount of calculation greatly, has reduced the realization cost.This invention is carried out smoothing processing to the vector field that estimates, the motion vector that can utilize adjacent block to because the mistake of the estimation that illumination or noise cause revise, improved the accuracy of estimation, and made motion vector field describe the real motion of object better.
Description of drawings
Fig. 1 is a schematic diagram of piece match search order;
Fig. 2 is another schematic diagram of piece match search order;
Fig. 3 is the another schematic diagram of piece match search order;
Fig. 4 is a candidate vector schematic diagram under the search order among Fig. 2;
Fig. 5 is another candidate vector schematic diagram under the search order among Fig. 2;
Fig. 6 enlarges the schematic diagram that block boundary carries out sub-sampling;
Fig. 7 is the schematic diagram of the motion vector field before a kind of medium filtering;
Fig. 8 is the schematic diagram of the motion vector field behind a kind of medium filtering.
Embodiment
Fig. 1 be traditional employing from left to right from top to bottom order the piece of present frame/field carried out the piece match search come estimated motion vector.
Fig. 2 adopts from top to bottom alternately from left to right and order is from right to left carried out the piece match search to the piece of present frame/field and come estimated motion vector.
Fig. 3 be adopt from around to the order of central authorities the piece of present frame/field is carried out the piece match search and comes estimated motion vector.
Fig. 4 has illustrated choosing of a kind of candidate vector under Fig. 2 search order.The piece of black is represented the current pixel piece among the figure, and the motion vector that is decorated with the piece of oblique line is a candidate vector.For the block of pixels of from left to right searching for, the motion vector of a piece on the motion vector that three candidate vectors are respectively the positive left adjacent blocks of current block, the motion vector that current block is just being gone up adjacent block, the current block bottom right 45 degree diagonals.
Fig. 5 has illustrated choosing of another kind of candidate vector under Fig. 2 search order.The piece of black is represented the current pixel piece among the figure, and the motion vector that is decorated with the piece of oblique line is a candidate vector.For the block of pixels of searching for from right to left, the motion vector of a piece on the motion vector that three candidate vectors are respectively the positive right adjacent blocks of current block, the motion vector that current block is just being gone up adjacent block, the current block lower-left 45 degree diagonals.
Based on piece match search selective sequential candidate vector, the motion vector of the piece of having handled before can utilizing, this can accelerate the speed of search so that the starting point of motion estimation search is more accurate, has reduced the amount of calculation of search; The candidate vector of selecting based on search order can utilize the temporal correlation of motion vector field better, more fully, makes that the performance of estimation is average in spatial distributions, improves visual effect.
Fig. 6 has illustrated that the expansion block boundary carries out sub-sampling.The piece that heavy line surrounds is original piece, and what thick dashed line surrounded is the piece that enlarges the border.is an original pixels, and ■ is the pixel that is used for motion estimation error calculating that sub-sampling is selected.On the piece that enlarges, carry out estimation, can make the result of estimation more accurate, and can reduce blocking effect; Sub-sampling can reduce the pixel that is used for motion estimation error calculating, reduces amount of calculation, and the sampling structure of sub-sampling can have multiple choices.
Fig. 7 has illustrated the motion vector field before the medium filtering.The heavy line area surrounded is exactly the zone of carrying out 2 dimension 3*3 medium filterings among the figure.The motion vector of current block is because of reasons such as illumination or noises as can be seen, and its motion vector is incorrect.
Fig. 8 has illustrated the motion vector field behind the medium filtering.Medium filtering is according to the spatial coherence of motion vector field as can be seen, utilizes the vector of contiguous block to correct the motion vector of current block, improved the performance of estimation, and made motion vector can describe the real motion of object better.
Concrete grammar of the present invention at first be according to from top to bottom alternately from left to right with the motion vector (as Fig. 4 or Fig. 5) of the piece of the motion vector of treated adjacent with the current block level and vertical adjacent two pieces of piece match search selective sequential from right to left, undressed current block 45 degree diagonals as tentatively definite candidate vector; To the motion vector of two in the preliminary candidate vector of the determining pieces of having handled according to the order of match search according to the selective sequential that replaces one of them and in the vector space of a preset range, select a vector randomly, this vector is added on the preliminary candidate vector of determining that those selected needs upgrade, forms the vector after upgrading.Other two that do not upgrade in vector after this upgrades and the preliminary candidate vector of determining constitute one group of candidate vector together.To each candidate vector, calculate the error of the reference pixel block of current pixel piece and its coupling then, calculate the error of same position block of pixels in current pixel piece and the reference frame simultaneously, the vector with minimum error values is defined as motion vector.Error amount can obtain by the SAD that calculates these pixels, the SAD minimum be confirmed as motion vector.2 dimension medium filterings that the level and the vertical component of the block motion vector that estimates adopted 3*3 respectively at last, if the motion vector that the horizontal component of medium filtering output and vertical component constitute is not any one that participates in 8 current block adjacent block motion vectors of medium filtering, the motion vector of getting original current block is as output, otherwise the result who gets medium filtering is as output, and the vector field before and after medium filtering is handled as shown in Figure 7 and Figure 8.
Claims (8)
1, video motion estimating method is characterized in that this method may further comprise the steps:
(a) based on the order of piece match search, according near the current block several motion vector and adopt a kind of update mechanism to determine one group of candidate vector;
(b) to each described candidate vector, calculate the error of the reference pixel block of current pixel piece and its coupling, calculate the error of same position block of pixels in current pixel piece and the reference frame simultaneously, the vector of choosing above error amount minimum is as motion vector;
(c) motion vector that estimates is carried out the level and smooth vector field of medium filtering;
The method of described step (a) is at first according to the motion vector of two pieces of having handled of selective sequential of piece match search, one undressed motion vector, three vectors upgrade the preliminary candidate vector of determining according to a kind of update mechanism then as the preliminary candidate vector of determining altogether.
2, video motion estimating method as claimed in claim 1, it is characterized in that the method for the error of calculation in the step (b) is the border that enlarges the block of pixels of the block of pixels of pending present frame/field and corresponding reference frame, and it is carried out sub-sampling, to the piece pixel behind sub-sampling, calculate the error of the reference pixel block of current pixel piece and its coupling then.
3, video motion estimating method as claimed in claim 1 is characterized in that the error of calculation is exactly directly to calculate the error of the reference pixel block of current pixel piece and its coupling in the step (b).
4, video motion estimating method as claimed in claim 1, it is characterized in that step (c) is at first the level of block motion vector and vertical component to be adopted the 2 dimension medium filterings of 3*3 respectively, result according to medium filtering output handles then, if the motion vector that the horizontal component of medium filtering output and vertical component constitute is not any one that participates in 8 current block adjacent block motion vectors of medium filtering, the motion vector of getting original current block is as output, otherwise the result who gets medium filtering is as output.
5, video motion estimating method as claimed in claim 1, the order that it is characterized in that the piece match search can be from top to bottom from left to right, from top to bottom from left to right, from top to bottom from right to left, from top to bottom from right to left; From top to bottom alternately from left to right and from right to left from top to bottom alternately from left to right and from right to left; From around to central authorities.
6, video motion estimating method as claimed in claim 1 is characterized in that the adjacent and vertical adjacent piece of level that described two treated pieces are current block; Described one undressed is the piece of 45 degree diagonals of current block.
7, video motion estimating method as claimed in claim 1, it is characterized in that described update mechanism be exactly at first to the motion vector of two in the preliminary candidate vector of the determining pieces of having handled according to the order of match search according to the selective sequential that replaces one of them; Then in the vector space of a preset range, select a vector randomly and this vector is added on selected that candidate vector of tentatively determining that upgrades, form the vector after upgrading, other two that do not upgrade in vector after this upgrades and the preliminary candidate vector of determining constitute one group of candidate vector together.
8, as claim 2 or 3 described video motion estimating methods, the error amount method that it is characterized in that calculating the reference pixel block of current pixel piece and its coupling is at first to the pixel summation of current pixel piece respective number, then the pixel of reference frame block of pixels respective number is sued for peace, to above-mentioned two and ask absolute difference, just obtain motion estimation error again.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CNB031151337A CN1181691C (en) | 2003-01-24 | 2003-01-24 | Vidio motion estimation method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CNB031151337A CN1181691C (en) | 2003-01-24 | 2003-01-24 | Vidio motion estimation method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN1444406A CN1444406A (en) | 2003-09-24 |
| CN1181691C true CN1181691C (en) | 2004-12-22 |
Family
ID=27814824
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CNB031151337A Expired - Lifetime CN1181691C (en) | 2003-01-24 | 2003-01-24 | Vidio motion estimation method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN1181691C (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1975782B (en) * | 2005-11-30 | 2012-11-21 | 汤姆逊许可公司 | Method of emendation for attention trajectory in video content analysis |
Families Citing this family (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| KR100846780B1 (en) * | 2003-11-10 | 2008-07-16 | 삼성전자주식회사 | Motion vector derivation method and apparatus |
| KR100907847B1 (en) * | 2004-07-20 | 2009-07-14 | 퀄컴 인코포레이티드 | Method and apparatus for motion vector prediction in temporal video compression |
| CN100340116C (en) * | 2005-01-21 | 2007-09-26 | 浙江大学 | Motion estimating method with graded complexity |
| CN100366092C (en) * | 2005-04-08 | 2008-01-30 | 北京中星微电子有限公司 | Search method for video frequency encoding based on motion vector prediction |
| CN100455023C (en) * | 2006-01-25 | 2009-01-21 | 华为技术有限公司 | Genetic searching method |
| CN101662679B (en) * | 2008-08-26 | 2011-12-21 | 华为技术有限公司 | Method and device for filtering waves |
| US8755437B2 (en) * | 2011-03-17 | 2014-06-17 | Mediatek Inc. | Method and apparatus for derivation of spatial motion vector candidate and motion vector prediction candidate |
| KR101532665B1 (en) * | 2011-03-14 | 2015-07-09 | 미디어텍 인크. | Method and apparatus for deriving temporal motion vector prediction |
| FR2986395A1 (en) * | 2012-01-30 | 2013-08-02 | France Telecom | CODING AND DECODING BY PROGRESSIVE HERITAGE |
| CN102970541B (en) * | 2012-11-22 | 2015-05-27 | 深圳市海思半导体有限公司 | Image filtering method and device |
| US10291931B2 (en) * | 2016-10-13 | 2019-05-14 | Ati Technologies Ulc | Determining variance of a block of an image based on a motion vector for the block |
| CN112954331B (en) * | 2019-03-11 | 2022-07-29 | 杭州海康威视数字技术股份有限公司 | Encoding and decoding method, device and equipment |
| CN111711823B (en) * | 2020-06-30 | 2022-11-15 | Oppo广东移动通信有限公司 | Motion vector processing method and apparatus, electronic device, and storage medium |
-
2003
- 2003-01-24 CN CNB031151337A patent/CN1181691C/en not_active Expired - Lifetime
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1975782B (en) * | 2005-11-30 | 2012-11-21 | 汤姆逊许可公司 | Method of emendation for attention trajectory in video content analysis |
Also Published As
| Publication number | Publication date |
|---|---|
| CN1444406A (en) | 2003-09-24 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN1925614A (en) | Rapid movement estimating method | |
| CN1181691C (en) | Vidio motion estimation method | |
| JP2920210B2 (en) | Motion vector prediction method for moving images | |
| CN101378504B (en) | Method for estimating block matching motion of H.264 encode | |
| EP1359765B1 (en) | Fading estimation/compensation in video encoding | |
| CN1675848A (en) | Method and apparatus for performing high quality fast predictive motion search | |
| CN1293872A (en) | Motion vector extrapolation for transcoding video sequences | |
| CN1208972C (en) | Motion estimating method and apparatus utilizing split-pole search and image coding system | |
| EP1359764A2 (en) | Video encoding method with fading compensation | |
| JPH08307874A (en) | Video signal encoding device | |
| EP2283658A2 (en) | Apparatus and method for computationally efficient intra prediction in a video coder | |
| CN1232125C (en) | Method for motion estimation (me) through discrete cosine transform (dct) and an apparatus therefor | |
| US20030231711A1 (en) | Interlaced video motion estimation | |
| US8073054B2 (en) | Unit for and method of estimating a current motion vector | |
| US20110129012A1 (en) | Video Data Compression | |
| JPH10191360A (en) | Method for obtaining motion estimate vector and method for compressing moving image data by using the motion estimate vector | |
| WO2000034920A1 (en) | Motion vector estimation | |
| Chen | Efficient block matching algorithm for motion estimation | |
| CN1263309C (en) | Motion vector prediction method used for video coding | |
| JP2008544598A (en) | Alternate up and down motion vectors | |
| CN101365133B (en) | DCT domain interposing and rounding error compensation process | |
| CN104902256B (en) | A kind of binocular stereo image decoding method based on motion compensation | |
| EP1484925A2 (en) | Method and device for compressing image data | |
| CN1137211A (en) | Improved motion compensation method for use in image encoding system | |
| KR0185841B1 (en) | A motion extimator for a video coder |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C14 | Grant of patent or utility model | ||
| GR01 | Patent grant | ||
| C56 | Change in the name or address of the patentee | ||
| CP03 | Change of name, title or address |
Address after: No. 90, Wensanlu Road, Hangzhou, Zhejiang, Xihu District Patentee after: HANGZHOU NATIONALCHIP SCIENCE & TECHNOLOGY Co.,Ltd. Address before: No. 516, Hangzhou Road, seven Xixi Road, Zhejiang, China Patentee before: HANGZHOU GUOXIN SCIENCE AND TE |
|
| CX01 | Expiry of patent term | ||
| CX01 | Expiry of patent term |
Granted publication date: 20041222 |