CN102075760B - Quick movement estimation method and device - Google Patents
Quick movement estimation method and device Download PDFInfo
- Publication number
- CN102075760B CN102075760B CN 201110047343 CN201110047343A CN102075760B CN 102075760 B CN102075760 B CN 102075760B CN 201110047343 CN201110047343 CN 201110047343 CN 201110047343 A CN201110047343 A CN 201110047343A CN 102075760 B CN102075760 B CN 102075760B
- Authority
- CN
- China
- Prior art keywords
- search
- pixel
- current block
- point
- size
- 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.)
- Active
Links
Images
Landscapes
- Compression Or Coding Systems Of Tv Signals (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a quick movement estimation method which comprises the following steps: taking the median prediction point of an existing block, the central point of a block at same position with the existing block in a reference frame and the motion vector toward point of the block at the same position with the existing block in the reference frame as candidate search points; selecting an initial search point in the candidate search points; determining the position of a search window in accordance with the selected initial search point, and determining the size of the search window in accordance with the type of the existing block; and determining the matching block/or the motion vector of the existing block in the search window. In the method, the multiple candidate search points are preset, and the advantages of a quick movement estimation algorithm are used for reference; and by setting the different search windows and search step sizes corresponding to the existing blocks with the different types, search accuracy is ensured, the total number of the points required to be searched can be effectively reduced at the same time, thus the method is more beneficial to realizing based on hardware.
Description
[technical field]
The present invention relates to field of video encoding, particularly be applicable to hard-wired rapid motion estimating method and device.
[background technology]
Video coding can be divided into video lossless coding and video compression coding two big classes.In DTV and video communication etc. were used, because the restriction of the network bandwidth and what the requirement of data volume size was generally used is video compression coding, H.261/263/264 international standard relatively more commonly used had, MPEG-1/2/4 etc.Video compression coding is generally taked based on piece or is claimed the macroblock encoding mode, promptly is divided into the two field picture in the video rectangular block of a plurality of non-overlapping copies, afterwards these pieces is encoded.
Because there is very strong correlation in time in video sequence image; Adopt estimation and motion compensation technique can eliminate time redundancy to improve code efficiency, for example in the video conference, background possibly be constant; And have only the personage, or personage's limbs move on picture.So; In adjacent two two field pictures in video (or a certain frame before present frame and the present frame); The piece possible position that shows same content is different, that is to say that certain displacement has taken place to be actually the piece that shows same content for the part that moves on the picture, certainly; Certain variation also possibly take place in this piece content displayed in two frames of front and back; Here claim that this piece in one frame of back is a current block, this piece in the former frame (or a certain frame before the said present frame, can be referred to as reference frame) is a match block.Research shows, in a video, has a lot of such pieces in present frame and the reference frame; Be that interframe has correlation; So, on the basis of reference frame coding, present frame needn't be encoded as reference frame is whole; Because there are a lot of identical data messages in present frame and the reference frame, promptly interframe is redundant.Common way is, deducts match block with current block, obtain between surplus and two displacement (also claiming motion vector) afterwards, obtain surplus is encoded with displacement data information, to carry out operations such as follow-up transmission or storage.Like this, when decoding,, just can recover the image of current block according to the coded message of match block and surplus and displacement information of current block.Obviously, in this course, it is little a lot of that surplus coding with displacement information of current block encoded than current block fully, and as far as a two field picture, the coding of back one frame is little more a lot of than the coding of former frame.
In the actual coding process, the process of from reference frame, seeking the match block of current block is exactly so-called " estimation ".Because current block and match block are not necessarily identical, so in the process of estimation, need in reference frame, find the piece similar as far as possible with current block.This moment two pieces similarity degree can by of the prior art certain estimate accurate value and calculate, for example absolute difference and (SAD), mean square error (MSE) and absolute error average (MAD) or the like.One two field picture is divided into non-overlapped rectangular block; It generally can be the piece of 16*16 pixel; Absolute value sum (the Sum of Absolute Difference of the difference of the pixel in the 16*16 pixel of two pieces in calculating present frame and the reference frame on each same position; SAD), can be expressed as with formula:
Wherein, i, j is remarked pixel transverse axis sequence number and longitudinal axis sequence number respectively, and x representes current block, and y representes a piece in the reference frame image.
The sad value of gained is more little, explains that these two piece matching degrees are high more, and promptly y possibly be the match block of x more.When calculating the match block of current block in reference frame, need some pieces in the selected reference frame to do calculating, be referred to as candidate blocks here.After in candidate blocks, finding match block, can obtain the displacement vector between reference block and the match block, i.e. motion vector.Foregoing can be with reference to figure 1.Among Fig. 1, the current block in the present frame is x, and y is the candidate blocks in the reference frame.After definite y is the match block of x, obtain x and y between motion vector MV, thereby accomplish the estimation of current block, subsequent treatment such as can encode afterwards.In this process, which piece in the reference frame is carried out the computing of match block as candidate blocks, a lot of different implementations are arranged, these implementations respectively have good and bad point.
Realize that in the prior art said process has based on the software realization with based on two big types of hardware realizations; Being based on hard-wired advantage can parallel computation, is not limited to the weakness according to flow performing that realizes based on software, in case but be based on hard-wired shortcoming and be that internal processes is set and finish; Follow-up very difficult modification; Even if be can parallel computation based on hard-wired advantage, but the calculation process on each road is all set follow-up can't the modification.Common fast motion estimation algorithm all is near selected point, to carry out search with good conditionsi, and these points spatially are discontinuous, and such characteristics cause realizing comparatively difficulty based on hardware, are more conducive to realize based on software.So in general; Based on hard-wired fast motion estimation algorithm all is the full-search algorithm that in the search window of certain limit, carries out; Promptly all pixel exhaustive search in this search window one time, but such technical scheme has increased amount of calculation undoubtedly, has expended a large amount of resources.
Therefore, demand urgently proposing a kind ofly can overcoming above-mentioned shortcoming, being applicable to hard-wired novel rapid motion estimating method.
[summary of the invention]
The purpose of this part is to summarize some aspects of embodiments of the invention and briefly introduces some preferred embodiments.In this part and the application's specification digest and denomination of invention, may do a little simplification or omit avoiding the making purpose of this part, specification digest and denomination of invention fuzzy, and this simplification or omit and can not be used to limit scope of the present invention.
One object of the present invention is to provide a kind of new rapid motion estimating method, is applicable to the hardware realization.
Another object of the present invention is to provide a kind of new fast motion estimating apparatus.
In order to reach the object of the invention; The present invention provides a kind of rapid motion estimating method, and said method comprises: with the motion vector points point of the central point of the median prediction point of current block, current block same position piece in reference frame and current block same position piece in reference frame as the candidate search point; In all candidate search points, select an initial search point; According to the position of the definite search window of the initial search point of selecting, and confirm the size of search window according to the type of current block; With match block and/or the motion vector of in said search window, confirming current block.
Further, select the method for initial search point to comprise: to establish candidate blocks respectively with said candidate search point; More said candidate blocks and said current block; Will with the minimum pairing candidate search point of candidate blocks of said current block matching error as initial search point.
Further, calculate the SAD between each candidate blocks and said current block respectively, select initial search point, also promptly select the minimum candidate search point of SAD as initial search point according to SAD.
Further, said type according to current block confirms that the size of search window is meant: current block is divided into the N class from small to large according to pixel, and every type of correspondence is scheduled to the search window of size or type, and said N is the integer greater than 0.
Further, said N is 3, and wherein the size of first kind current block is the 4*4 pixel; The size of second type of current block is 8*4 pixel, 4*8 pixel and/or 8*8 pixel; With the size of the 3rd type of current block be 8*16 pixel, 16*8 pixel and/or 16*16 pixel.
Further, the search window of said predetermined size comprises: first kind search window is the rectangular search window of 5*5 pixel, and step-size in search is 1 pixel in it; Second type of search window is that the outer length of side is that 15*15 pixel, the interior length of side are the three-back-shaped search window of 5*5 pixel, and step-size in search is 2 pixels in it; With the 3rd type of search window be that the outer length of side is that 32*32 pixel, the interior length of side are the three-back-shaped search window of 15*15 pixel, step-size in search is 4 pixels in it.
Further, only in first kind search window, search for for first kind current block; Search in first kind search window and second type of search window for second type of current block; With search in first kind search window, second type of search window and the 3rd type of search window for the 3rd type of current block.
Further, for same current block, adopt parallel search method when in different search window, searching for its match block.
Further, said candidate search point also comprises:
Apart from the step-length up and down of the median prediction point of current block be 1 pixel four points and
Step-length up and down apart from the central point of current block same position piece in reference frame is four points of 1 pixel.
Further, said position according to the definite search window of the initial search point of selecting is meant: the initial search point that will select is as the central point of said search window.
Further; Selecting initial search point and confirming in the process of match block and/or motion vector of current block; All be to be used as matching criterior according to predetermined matching criterior, said predetermined matching criterior is that absolute value sum according to the difference of the pixel on each same position between two pieces is as matching criterior.
According to a further aspect in the invention, the present invention also provides a kind of fast motion estimating apparatus, and said device comprises: starting point is selected module, in all candidate search points of current block, selects an initial search point; The search determination module is searched for according in type selecting first search module, second search module and the 3rd search module of current block one or more; First search module is set up the rectangular search window of 5*5 pixel for the center and is that the match block that the mode of 1 pixel is carried out current block is searched for step-size in search according to the initial search point of selecting; Second search module is that 15*15 pixel, the interior length of side are the three-back-shaped search window of 5*5 pixel and are that the match block that the mode of 2 pixels is carried out current block is searched for step-size in search according to the initial search point of selecting for the center outer length of side of foundation; The 3rd search module is that 32*32 pixel, the interior length of side are the three-back-shaped search window of 15*15 pixel and are that the match block that the mode of 4 pixels is carried out current block is searched for step-size in search according to the initial search point of selecting for the center outer length of side of foundation; And comparison module, the match block and/or the motion vector of definite current block behind the Search Results in receiving said first search module, second search module and/or the 3rd search module.
Further, all candidate search points of said current block comprise: the median prediction point of current block and apart from the step-length up and down of this point be 1 pixel four points, current block same position piece in reference frame central point and be four points of 1 pixel and the motion vector points point of current block same position piece in reference frame apart from the step-length up and down of this point.
Further, it all is to be used as matching criterior according to predetermined matching criterior that said starting point is selected module and comparison module, and said predetermined matching criterior is that absolute value sum according to the difference of the pixel on each same position between two pieces is as matching criterior.
Further, when the size of current block was the 4*4 pixel, said search determination module was selected the search of first search module;
When the size of current block was 8*4 pixel, 4*8 pixel and/or 8*8 pixel, said search determination module was selected first search module and the search of second search module; With
When the size of current block was 8*16 pixel, 16*8 pixel and/or 16*16 pixel, said search determination module was selected first search module, second search module and the search of the 3rd search module.
Compared with prior art, the present invention has used for reference the advantage of existing fast motion estimation algorithm through predetermined a plurality of candidate search points; Through setting dissimilar current blocks corresponding different search window and step-size in search, when guaranteeing search precision, reduced the overall quantity that needs the point of search effectively, make the present invention be more conducive to realize based on hardware; For same current block, can also adopt parallel search method when in different search window, searching for its match block, make full use of based on hard-wired advantage.
[description of drawings]
In conjunction with reference to accompanying drawing and ensuing detailed description, the present invention will be more readily understood, the structure member that wherein same Reference numeral is corresponding same, wherein:
Fig. 1 is the location diagram between current block, match block and the motion vector in the prior art;
Fig. 2 is the method flow diagram of the rapid motion estimating method in one embodiment of the present of invention;
Fig. 3 is boundary block and the location diagram between current block and its adjacent piece in one embodiment of the present of invention;
Fig. 4 is the position view of the candidate search point in one embodiment of the present of invention;
Fig. 5 is the structural representation of the search window in one embodiment of the present of invention; With
Fig. 6 is the structural representation of the fast motion estimating apparatus in one embodiment of the present of invention.
[embodiment]
For make above-mentioned purpose of the present invention, feature and advantage can be more obviously understandable, below in conjunction with accompanying drawing and embodiment the present invention done further detailed explanation.
The present invention provides rapid motion estimating method and the device in the video coding, and main feature is to have avoided the shortcoming that exists hardware such as irregular address jump to be difficult to realize in traditional rapid motion estimating method.In general, the motion estimation process for piece can be divided into two steps: the position of at first confirming initial search point in reference frame; Be that search window is set up at the center with this initial search point then, carry out the search of the match block of current block.One of emphasis of the present invention and bright spot be exactly in these two steps at least one step for the benefit of hardware realize and made corresponding improvement.
Please refer to Fig. 2, it shows the method flow diagram of the rapid motion estimating method 200 in one embodiment of the present of invention.Said rapid motion estimating method 200 comprises:
When step 202, current block are non-boundary block, with the motion vector points point of the central point of the median prediction point of current block, current block same position piece in reference frame and current block same position piece in reference as the candidate search point.
The main candidate search point of confirming in the reference frame in this step; Be divided into the rectangular block of a plurality of non-overlapping copies when a two field picture after; Its most upper left drift angle piece is referred to as upper left boundary block, and the piece of its Far Left except that the most upper left drift angle piece is referred to as the left margin piece, correspondingly; Its topmost the piece except that the most upper left drift angle piece be referred to as the coboundary piece, shown in Fig. 3 A.When current block was non-boundary block, the adjacent piece of current block comprised left piece, go up piece and upper right, shown in Fig. 3 B.Usually these three adjacent pieces that utilize current block in the conventional method carry out median prediction to the motion vector of current block; Be about to current block left piece, go up piece and upper right the vector of horizontal component median and vertical component median formation of motion vector as the motion vectors of current block, set up search window with the median prediction point that this motion vectors is pointed to as initial search point then in reference frame.The principle of this median prediction is the correlation of utilizing between current block and the adjacent block, and prediction generally is comparatively accurate and can reduce computational complexity.But when not belonging to the situation of an object in fact such as current block and adjacent block, this median prediction can form a misleading on the contrary.So in this step not only with median prediction point as the candidate search point, simultaneously with the motion vector points point of the central point of current block same position piece in reference frame and current block same position piece in reference frame also as the candidate search point.Said same position piece also is referred to as corresponding blocks, is and big or small identical macro block identical with the position of current block in reference frame.In one embodiment; Said candidate search point comprises that also the step-length up and down apart from the median prediction point of current block is four points of 1 pixel and is four points of 1 pixel apart from the step-length up and down of the central point of current block same position piece in reference frame.Just in this step, said candidate search point can comprise 11 points, to guarantee the accuracy of initial search point.
As shown in Figure 4, in the reference frame of current block, confirmed the candidate search point, wherein the A point carries out the median prediction point that the motion vectors MV1 after the median prediction points to for the adjacent piece that utilizes current block; The B point is the central point of the same position piece of current block in reference frame; The C point is the sensing point of the motion vector MV2 of the corresponding blocks of current block in reference frame.Simultaneously the candidate search point can also comprise that the step-length up and down of distance A point is four points of 1 pixel and is four points of 1 pixel apart from the step-length up and down that B is ordered.To mating with current block respectively as the candidate blocks of central point, select matching error candidate search point hour as initial search point with said candidate search point.Particularly, can be to be matching criterior with SAD, calculate the SAD of current block and candidate blocks, select initial search point according to SAD, pairing candidate search point is as next step initial search point under the situation of SAD minimum.
Search window as its central point, therefore according to the initial search point of selecting in the step 204, can be confirmed the position of search window with initial search point, in this step, also need confirm the size of said search window.Because in the different video compression coding standards; Partitioned mode for current block is different; Such as in standard H.264, the partitioned mode of said current block is: 16*16 pixel, 16*8 pixel, 8*16 pixel, 8*8 pixel, 8*4 pixel, 4*8 pixel and 4*4 pixel.So, also should take different search window size with the minimizing computational complexity in order to improve the search accuracy for the current block of different sizes.In one embodiment, current block is divided into the N class from small to large according to pixel, every type of correspondence is scheduled to the search window of size or type, and said N is the integer greater than 0.For example, can said current block be divided into 3 types from small to large according to pixel, wherein for the partitioned mode in the standard H.264
The size of first kind current block is the 4*4 pixel;
The size of second type of current block is 8*4 pixel, 4*8 pixel and/or 8*8 pixel; With
The size of the 3rd type of current block is 8*16 pixel, 16*8 pixel and/or 16*16 pixel.
Then accordingly, the search window of the predetermined size of definition is following:
First kind search window is that step-size in search is the rectangular search window of the 5*5 pixel of 1 pixel;
Second type of search window is that step-size in search is that the outer length of side of 2 pixels is that 15*15 pixel, the interior length of side are the three-back-shaped search window of 5*5 pixel; With
The 3rd type of search window is that step-size in search is that the outer length of side of 4 pixels is that 32*32 pixel, the interior length of side are the three-back-shaped search window of 15*15 pixel.Said search window can be with reference to figure 5.
When concrete search, only in first kind search window, search for for first kind current block; Search in first kind search window and second type of search window for second type of current block; With search in first kind search window, second type of search window and the 3rd type of search window for the 3rd type of current block.Take such mode to be because inventor's research shows, when a pixel surpassed 5 pixels apart from the distance of current pixel point, the correlation between them was very low.Usually when taking exercises estimation, the probability that in the distance of 5*5 pixel, finds the optimum Match pixel is more than 90%.Certainly the predetermined size of mode classification in this step and search window maybe be different with the technical application personnel because of different enforcement environment, but if essence according to the invention, all should be as protection scope of the present invention.
Stipulate different step-size in searchs for different search window, can effectively reduce the overall quantity that needs the point of search like this, reduced computational complexity.The position and the number of the search point in each search window are all fixed simultaneously, and are confirmable in advance, so do not exist hardware need handle problems such as address jump in this step.On the other hand, for same current block, can also adopt parallel search method when in different search window, searching for its match block.The current block that for example is of a size of the 8*8 pixel should all be searched in first kind search window and second type of search window, and can adopt parallel search method to make full use of based on hard-wired advantage this moment.
Said predetermined matching criterior can also adopt absolute difference and SAD as matching criterior.Promptly in search procedure, the candidate blocks of selecting minimum SAD is as match block.When adopting parallel search method, in each search window, search for finish after, can also compare the local optimum point in each search window, the candidate blocks of finally selecting minimum SAD calculates the motion vector of current block simultaneously as match block.
After estimation finishes, can utilize said motion vector that current block is carried out next code and handle.
Will be appreciated that; The matching criterior of taking among the present invention is not limited to absolute difference and SAD, and this is a kind of, and said matching criterior can also adopt absolute error and STAD, squared difference and SSD and mean absolute difference MAD or the like after rate distortion coefficients R DO, the Hadleman conversion.The step-size in search of the predetermined size of the classification of current block, search window and search window all might be according to the different environment of implementing with the technical application personnel and maybe be different in step 206.
The present invention proposes a kind of fast motion estimating apparatus simultaneously, please refer to Fig. 6, and it shows the structural representation of the fast motion estimating apparatus 600 in one embodiment of the present of invention.Said fast motion estimating apparatus 600 comprises that starting point selects module 601, search determination module 602, first search module 603, second search module 604, the 3rd search module 605 and comparison module 606
Starting point selects module 601 in all candidate search points of current block, to select initial search point.The candidate search point of said current block comprises: the median prediction point of current block and apart from the step-length up and down of this point be 1 pixel four points, current block same position piece in reference frame central point and be four points of 1 pixel and the motion vector points point of current block same position piece in reference frame apart from the step-length up and down of this point.
The 3rd search module 605 is that 32*32 pixel, the interior length of side are the three-back-shaped search window of 15*15 pixel and are that the match block that the mode of 4 pixels is carried out current block is searched for step-size in search according to the initial search point of selecting for the center outer length of side of foundation; With
Confirm the match block and/or the motion vector of current block behind the Search Results of comparison module 606 in receiving said first search module 603, second search module 604 and/or the 3rd search module 605.
Said starting point selects module 601 to confirm best initial search point according to predetermined matching criterior; Said comparison module 606 confirms that according to predetermined matching criterior the match block of current block, said predetermined matching criterior are that absolute value sum according to the difference of the pixel on each same position between two pieces is as matching criterior.
Above-mentioned explanation has fully disclosed embodiment of the present invention.It is pointed out that any change that technical staff's specific embodiments of the invention of being familiar with this field is done does not all break away from the scope of claims of the present invention.Correspondingly, the scope of claim of the present invention also is not limited only to said embodiment.
Claims (10)
1. rapid motion estimating method is characterized in that it comprises:
With the motion vector points point of the central point of the median prediction point of current block, the same position piece of current block in reference frame and current block same position piece in reference frame as the candidate search point;
In all candidate search points, select initial search point;
According to the position of the definite search window of the initial search point of selecting, and confirm the size of search window according to the size of current block; With
In said search window, confirm the match block and/or the motion vector of current block,
Said size according to current block confirms that the size of search window is meant:
Current block is divided three classes according to pixel from small to large, the search window of every type of predetermined size of correspondence, wherein
The size of first kind current block is the 4*4 pixel;
The size of second type of current block is 8*4 pixel, 4*8 pixel and/or 8*8 pixel; With
The size of the 3rd type of current block is 8*16 pixel, 16*8 pixel and/or 16*16 pixel,
The search window of said predetermined size comprises:
First kind search window is the rectangular search window of 5*5 pixel, and step-size in search is 1 pixel in it;
Second type of search window is that the outer length of side is that 15*15 pixel, the interior length of side are the three-back-shaped search window of 5*5 pixel, and step-size in search is 2 pixels in it; With
The 3rd type of search window is that the outer length of side is that 32*32 pixel, the interior length of side are the three-back-shaped search window of 15*15 pixel, and step-size in search is 4 pixels in it,
Only in first kind search window, search for for first kind current block;
Search in first kind search window and second type of search window for second type of current block; With
Search in first kind search window, second type of search window and the 3rd type of search window for the 3rd type of current block.
2. method according to claim 1 is characterized in that, selects the method for initial search point to comprise: to establish candidate blocks respectively with said candidate search point; More said candidate blocks and said current block; Will with the minimum pairing candidate search point of candidate blocks of said current block matching error as initial search point.
3. method according to claim 2 is characterized in that, calculates the SAD between each candidate blocks and said current block respectively, selects initial search point according to SAD, also promptly selects the minimum candidate search point of SAD as initial search point.
4. method according to claim 1 is characterized in that, for same current block, adopts parallel search method when in different search window, searching for its match block.
5. method according to claim 1 is characterized in that, said candidate search point also comprises:
Apart from the step-length up and down of the median prediction point of current block be 1 pixel four points and
Step-length up and down apart from the central point of current block same position piece in reference frame is four points of 1 pixel.
6. method according to claim 1 is characterized in that, said position according to the definite search window of the initial search point of selecting is meant: the initial search point that will select is as the central point of said search window.
7. method according to claim 1; It is characterized in that; Selecting initial search point and confirming in the process of match block and/or motion vector of current block; All be to be used as matching criterior according to predetermined matching criterior, said predetermined matching criterior is that absolute value sum according to the difference of the pixel on each same position between two pieces is as matching criterior.
8. fast motion estimating apparatus is characterized in that it comprises:
Starting point is selected module, in all candidate search points of current block, selects an initial search point;
The search determination module selects in first search module, second search module and the 3rd search module one or more to search for according to the size of current block;
First search module is set up the rectangular search window of 5*5 pixel for the center and is that the match block that the mode of 1 pixel is carried out current block is searched for step-size in search according to the initial search point of selecting;
Second search module is that 15*15 pixel, the interior length of side are the three-back-shaped search window of 5*5 pixel and are that the match block that the mode of 2 pixels is carried out current block is searched for step-size in search according to the initial search point of selecting for the center outer length of side of foundation;
The 3rd search module is that 32*32 pixel, the interior length of side are the three-back-shaped search window of 15*15 pixel and are that the match block that the mode of 4 pixels is carried out current block is searched for step-size in search according to the initial search point of selecting for the center outer length of side of foundation; With
Comparison module, the match block and/or the motion vector of definite current block behind the Search Results in receiving said first search module, second search module and/or the 3rd search module,
When the size of current block was the 4*4 pixel, said search determination module was selected the search of first search module;
When the size of current block was 8*4 pixel, 4*8 pixel and/or 8*8 pixel, said search determination module was selected first search module and the search of second search module; With
When the size of current block was 8*16 pixel, 16*8 pixel and/or 16*16 pixel, said search determination module was selected first search module, second search module and the search of the 3rd search module.
9. device according to claim 8; It is characterized in that all candidate search points of said current block comprise: the median prediction point of current block and apart from the step-length up and down of this point be 1 pixel four points, current block same position piece in reference frame central point and be four points of 1 pixel and the motion vector points point of current block same position piece in reference frame apart from the step-length up and down of this point.
10. device according to claim 8; It is characterized in that; It all is to be used as matching criterior according to predetermined matching criterior that said starting point is selected module and comparison module, and said predetermined matching criterior is that absolute value sum according to the difference of the pixel on each same position between two pieces is as matching criterior.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN 201110047343 CN102075760B (en) | 2010-10-27 | 2011-02-28 | Quick movement estimation method and device |
Applications Claiming Priority (3)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201010520580.9 | 2010-10-27 | ||
| CN201010520580 | 2010-10-27 | ||
| CN 201110047343 CN102075760B (en) | 2010-10-27 | 2011-02-28 | Quick movement estimation method and device |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN102075760A CN102075760A (en) | 2011-05-25 |
| CN102075760B true CN102075760B (en) | 2012-11-21 |
Family
ID=44034082
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN 201110047343 Active CN102075760B (en) | 2010-10-27 | 2011-02-28 | Quick movement estimation method and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN102075760B (en) |
Families Citing this family (16)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102917216A (en) * | 2012-10-16 | 2013-02-06 | 深圳市融创天下科技股份有限公司 | Motion searching method and system and terminal equipment |
| CN103064973A (en) * | 2013-01-09 | 2013-04-24 | 华为技术有限公司 | Method and device for searching extreme values |
| CN103079072A (en) * | 2013-01-15 | 2013-05-01 | 清华大学 | Inter-frame prediction method, encoding equipment and decoding equipment |
| CN103974077B (en) * | 2014-05-14 | 2017-04-19 | 东南大学 | Quick integer motion estimation searching method used for H.264 coding |
| CN106331703B (en) | 2015-07-03 | 2020-09-08 | 华为技术有限公司 | Video encoding and decoding method, video encoding and decoding device |
| CN105007493B (en) * | 2015-07-17 | 2018-06-29 | 珠海市杰理科技股份有限公司 | Matching searching method and system based on motion search window |
| CN106375769B (en) * | 2016-08-31 | 2019-04-30 | 西安万像电子科技有限公司 | Characteristics of image searching method and device, storage medium and processor |
| 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 |
| CN108900846A (en) * | 2018-07-17 | 2018-11-27 | 珠海亿智电子科技有限公司 | A kind of the two-dimensional directional motion estimation hardware circuit and its method of Video coding |
| CN109561315B (en) * | 2018-10-31 | 2020-10-02 | 北京金山云网络技术有限公司 | Motion estimation method and device, electronic equipment and storage medium |
| KR102476057B1 (en) | 2019-09-04 | 2022-12-09 | 주식회사 윌러스표준기술연구소 | Method and apparatus for accelerating video encoding and decoding using IMU sensor data for cloud virtual reality |
| CN111598919B (en) * | 2020-04-22 | 2023-06-30 | Oppo广东移动通信有限公司 | Motion estimation method, motion estimation device, storage medium and electronic equipment |
| CN111800631B (en) * | 2020-06-22 | 2024-07-23 | 西安万像电子科技有限公司 | Data processing method and system |
| CN111784734B (en) * | 2020-07-17 | 2024-07-02 | Oppo广东移动通信有限公司 | Image processing method and device, storage medium and electronic equipment |
| CN116074533B (en) * | 2023-04-06 | 2023-08-22 | 湖南国科微电子股份有限公司 | Motion vector prediction method, system, electronic device and storage medium |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1333634A (en) * | 2001-01-12 | 2002-01-30 | 北京航空航天大学 | Quick video motion estimating method |
| CN1719901A (en) * | 2004-07-08 | 2006-01-11 | 三星电子株式会社 | Recording medium based on estimation multiresolution method and its program of storage execution |
| CN1816146A (en) * | 2005-02-03 | 2006-08-09 | 联想(北京)有限公司 | Method for estimating motion vector in video frequency data compression |
-
2011
- 2011-02-28 CN CN 201110047343 patent/CN102075760B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1333634A (en) * | 2001-01-12 | 2002-01-30 | 北京航空航天大学 | Quick video motion estimating method |
| CN1719901A (en) * | 2004-07-08 | 2006-01-11 | 三星电子株式会社 | Recording medium based on estimation multiresolution method and its program of storage execution |
| CN1816146A (en) * | 2005-02-03 | 2006-08-09 | 联想(北京)有限公司 | Method for estimating motion vector in video frequency data compression |
Non-Patent Citations (3)
| Title |
|---|
| 王伟等.基于可变分块尺寸的多模板运动估计算法.《自动化学报》.2009,第35卷(第1期),全文. * |
| 鲁小兵等.H.264运动估计搜索窗口的动态调整算法.《图像图形技术研究与应用2009》.2009 第3.1.2节 |
| 鲁小兵等.H.264运动估计搜索窗口的动态调整算法.《图像图形技术研究与应用2009》.2009,第3.1.2节,3.2节. * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN102075760A (en) | 2011-05-25 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN102075760B (en) | Quick movement estimation method and device | |
| CN100468982C (en) | Method and apparatus for performing high quality fast predictive motion search | |
| CN100539693C (en) | Rapid motion estimating method | |
| CN101600108B (en) | Joint estimation method for movement and parallax error in multi-view video coding | |
| JP5138048B2 (en) | Adaptive motion estimation | |
| CN118055253A (en) | Optical flow estimation for motion compensated prediction in video coding | |
| CN103079069A (en) | Decoding method | |
| CN101184233B (en) | CFRFS based digital video compressed encoding method | |
| US11115678B2 (en) | Diversified motion using multiple global motion models | |
| CN103327327B (en) | For the inter prediction encoding unit selection method of high-performance video coding HEVC | |
| CN107360433B (en) | Inter-frame prediction coding method and device | |
| KR20070033345A (en) | How to retrieve global motion vector | |
| CN112004097B (en) | Inter-frame prediction method, image processing apparatus, and computer-readable storage medium | |
| CN113709458B (en) | Displacement vector prediction method, device and equipment in video coding and decoding | |
| CN110545425A (en) | Inter-frame prediction method, terminal equipment and computer storage medium | |
| CN109688411B (en) | Video coding rate distortion cost estimation method and device | |
| CN104333758B (en) | The method and relevant apparatus of prediction technique and the detection pixel point of depth map | |
| US8837595B2 (en) | Motion estimation method | |
| CN101931739A (en) | Absolute error sum estimation system and method | |
| CN100385957C (en) | Kinematic vector predicting method | |
| CN104113756A (en) | Integer pixel motion estimation method suitable for H.264 video encoding and decoding | |
| US20210144364A1 (en) | Motion Field Estimation Based on Motion Trajectory Derivation | |
| KR20040035390A (en) | Method and apparatus for motion estimation using of adaptive search pattern for video sequence compression | |
| CN202931463U (en) | Characteristic block based video image stabilization device | |
| Mistry et al. | Logarithmic search for motion estimation |
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 | ||
| TR01 | Transfer of patent right |
Effective date of registration: 20170817 Address after: 226300 Tongzhou District, Nantong, Century Avenue, No. 6, layer 999, No. Patentee after: Jiangsu Dahai Intelligent System Co., Ltd. Address before: 214028 Jiangsu New District of Wuxi, Taihu international science and Technology Park Jia Qing 530 building 10 layer Patentee before: Wuxi Vimicro Co., Ltd. |
|
| TR01 | Transfer of patent right |