CN100385955C - Block displacement prognosticating method - Google Patents
Block displacement prognosticating method Download PDFInfo
- Publication number
- CN100385955C CN100385955C CNB2004100562645A CN200410056264A CN100385955C CN 100385955 C CN100385955 C CN 100385955C CN B2004100562645 A CNB2004100562645 A CN B2004100562645A CN 200410056264 A CN200410056264 A CN 200410056264A CN 100385955 C CN100385955 C CN 100385955C
- Authority
- CN
- China
- Prior art keywords
- search
- block
- motion vector
- seek operations
- searching
- 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)
Abstract
The present invention relates to a displacement prediction method for determining outcome displacement vector, which comprises: beginning with initial displacement vector, first searching operation is done to determine middle displacement vector, wherein the first searching operation searches according to a first searching principle; beginning with the middle displacement vector, second searching operation is done to determine result displacement vector, wherein the second searching operation searches according to a second searching principle. The first searching operation is rough searching operation, and the second searching operation is a fine searching operation.
Description
Technical field
The present invention relates to a kind of blurring predictor method, particularly relate to and a kind ofly utilize the known block in two frames to decide in the blurring predictor method of the motion vector of a block.
Background technology
In the coded system that digitized video is handled, are its main purposes with a large amount of compressions of the capacity of data, and include some coding kenels usually in the specification that general video signal is handled, to adapt to the coding demand of various video signal datas.With moving picture coding expert group (Moving Picture Coding Experts Group, MPEG-2 specification MPEG) is an example, it mainly is to eliminate on the video pictures space and temporal redundant information, to realize the video files compression effects.
The general video signal of being play, be a succession of continuous images sequence in fact, yet because there is a kind of existing picture of the persistence of vision on the human vision, so can produce the illusion of the image that links up, and in this kind video signal, because the time interval between picture is very little, so several indifferences between the picture that faces mutually are the change in location of picture material mostly.Therefore, in the MPEG-2 specification,, realize compressing the purpose of video signal, and displacement pre-estimating technology wherein (motion estimation) is a considerable ring in the mode of the redundant information that similitude was caused between picture and picture on the elimination time shaft.In order to remove aforementioned temporal redundant information, in the MPEG-2 specification, used the video signal compress technique of bit shift compensation (motion compensation).Before carrying out bit shift compensation, the block of at first a frame being divided into many specific big or small pixels (or claims macro zone block, macroblock); Then, with a block is that unit is as a target block, and in a reference frame (reference frame), look for a reference block approximate, and this target macro block to this coordinate difference with reference to position between macro zone block is recorded as a motion vector (motion vector) with this target block.
The digitized video processing mode of generally knowing is in the computing of seeking motion vector, usually use considerable memory size and time, mode with overall search (exhaustive search) in a search area (search range) of being scheduled to finds and another approximate block of original block, such motion vector is sought computing may not only increase the volume of hardware on real the work, even can procrastinate the task performance of hardware.Though existing many documents are devoted to improve this phenomenon, L.Liu for example, E.Feig, " Ablock-based gradient descent search algorithm for block motionestimation in video coding; " IEEE Trans.on Circuits and Systems forVideo Technology, vol.6, no.4, pp.419-422, P.I.Hosur, K.K.Ma, " Motion vector field adaptive fast motion estimation; " SecondInternational Conference on Information Communication and SignalProcessing, Singapore, 7-10 Dec ' 99, and Y.Nie, K.K.Ma, " Adaptiverood pattern search for fast block-matching motion estimation; " IEEETrans.on Image Processing, vol.11, no.12, pp.1442-1449, Dec ' 02, and waiting all has disclosure in the document, but still needs further to make great efforts.
Summary of the invention
Therefore, one of main purpose of the present invention provides a kind of displacement predictor method that reduces the operation time of motion vector.
According to embodiments of the invention, disclosed a kind of displacement predictor method, with deciding a motion vector as a result, this displacement predictor method includes: since initial displacement vector, carry out one first seek operations, with motion vector in the middle of determining, wherein this first seek operations is searched according to one first search principle; And, carry out one second seek operations from this centre motion vector, to determine this motion vector as a result, wherein this second seek operations is searched according to one second search principle.Described first searches principle includes the initial search direction that direction is searched in one of selection as this first seek operations in two search directions, wherein these two to search direction vertical mutually or be a fixed angle, and, search the enterprising line search of direction at these two in turn from this initial direction of searching
According to embodiments of the invention, also disclosed a kind of displacement predictor method, with deciding one first motion vector, this displacement predictor method includes: search two and select one as an initial direction of searching in the direction; These initial a plurality of blocks of searching on the direction are carried out seek operations, determine one first block with an error function; A plurality of blocks on the other direction of this initial search direction are carried out seek operations, determine one second block with this error function or another error function; And when this first block and this second block are same block, will be made as this first motion vector corresponding to the motion vector of this block.
Description of drawings
Fig. 1 is the schematic diagram according to the handled block of displacement predictor method of one embodiment of the invention.
Fig. 2 is the operational flowchart according to the displacement predictor method of one embodiment of the invention.
Fig. 3 is the operational flowchart of searching according to the displacement adaptability circulation of one embodiment of the invention.
Fig. 4 is the operational flowchart of searching on a large scale according to one embodiment of the invention.
Fig. 5 is the operational flowchart of searching among a small circle according to one embodiment of the invention.
Fig. 6 carries out a schematic diagram of searching on a large scale according to the displacement predictor method of one embodiment of the invention.
Fig. 7 carries out a schematic diagram of searching among a small circle according to the displacement predictor method of one embodiment of the invention.
The reference numeral explanation
| |
210，220 |
| Block | 202，204，206，208，212，214，216，218，222，224， 226，228，232，234，236，238，242，244，602，604， 606，608，612，712，832，834，836，838 |
| |
622，624，626，628，632，702，712，714，716，718， 732，734，736，738，740 |
Embodiment
See also Fig. 1, Fig. 1 is the schematic diagram according to the handled block of displacement predictor method of one embodiment of the invention, the pending block 202 (currentblock) in the pending frame 220 wherein, block 204,206,208 (surrounding block) (be respectively top block T, go up right-hand block TR, and left block L) on every side, and the size of other block 212,214,216,218,222 etc. is 8 * 8, that is in the present embodiment, each block is a square, and is made of 64 pixels.Similarly, for reference frame 210, the last block (previousblock) 234 that corresponds to pending block 202 with and also corresponding 8 * 8 size and being constituted of peripheral region piece 224,226,228,232,236,238,242,244 by 64 pixels, please note, in the displacement pre-estimating technology, the block size is not to be defined as 8 * 8, and can utilize the block (as 16 * 16) of other size to come the displacement calculating vector.As shown in Figure 1, last block 234 is identical corresponding to the display position of this display frame picture with pending frame 220 corresponding to the display position of a display frame in the reference frame 210.
See also Fig. 2, Fig. 2 is the operational flowchart according to the displacement predictor method of one embodiment of the invention.In the present embodiment, this displacement predictor method includes following steps at the running of these pending block 202 displacement calculating vectors:
Step 100: operation is estimated in the beginning displacement;
Step 102: determine initial set (the initial search set) S that searches.In the present embodiment, initial search include in the S set pending block 202 with and a plurality of around the motion vector of block, described herein a plurality of around blocks include top block 204, go up right-hand block 206, left block 208 and last block 234.In other embodiments, initially search the set that S set also can be the motion vector that comprises the different blocks combination;
Step 104:, calculate an error function (costfunction) at each element in the initial search S set.In the present embodiment, this error function be to use the absolute error summation (sum ofabsolute difference, mode SAD) realizes, knows that error function can also be realized by alternate manner yet extensively know as those skilled in the art;
Step 106: with in the step 104 according to the functional value reckling that this error function calculated, compare with one first critical value, if this smallest error function value is less than this first critical value, then execution in step 108, otherwise execution in step 110.In the present embodiment, this first critical value is by obtaining in the data of before having finished search, is made as this last block 234 and the error function value reckling of block 224～244 on every side thereof herein, yet in other embodiments, this first critical value also can be other numerical value, and obtains according to distinct principle;
Step 108: the motion vector that will have the block of smallest error function value promptly is made as the motion vector of search result;
Step 110: carry out displacement adaptability circulation and search (motion adaptive loopsearch), the detailed description about this circulation is searched please refer to following Fig. 3 and related description thereof;
Step 112: finish displacement and estimate operation.
Next see also Fig. 3, Fig. 3 is the operational flowchart of searching according to the disclosed displacement adaptability circulation of one embodiment of the invention.In the present embodiment, this flow process is corresponding to the step 110 among Fig. 2, and includes following steps:
Step 300: the circulation of beginning displacement adaptability is searched;
Step 302: the diversity factor of displacement calculating vector (variation), so-called herein diversity factor are represented the total difference degree of the motion vector of block on every side of pending block.The motion vector of supposing top block 204 is expressed as (MV
XT, MV
YT), the motion vector of going up right-hand block 206 is expressed as (MV
XTR, MV
YTR), and the motion vector of left block 208 is expressed as (MV
XL, MV
YL), then in the present embodiment, diversity factor can be complied with down routine equation and calculate:
medMV
X＝median(MV
XT，MV
XTR，MV
XL) (1)
medMV
Y＝median(MV
YT，MV
YTR，MV
YL) (2)
variation＝varMV
X+varMV
Y (5)
MedMV wherein
XAnd medMV
YBe the figure place among on X component and the Y component of block around described, varMV
XAnd varMV
YBe described around the variance (variance) of block on X component and Y component, and diversity factor (variation) be the summation of above-mentioned two variances in the present embodiment, still also other operation result of two variances for this reason in other embodiments;
Step 304: above-mentioned diversity factor is compared with one second critical value, and less than this second critical value, then execution in step 308 as if this diversity factor, otherwise execution in step 306.Carry out standardization computing (normalization) as if the diversity factor that aforementioned calculation is come out herein, then this second critical value can be made as 1, but also can be according to other numerical value of practical application substitution;
Step 306: carry out one and search on a large scale, it is a rough seek operations (coarse search), is used for dwindling more quickly the scope of search.About this detailed description of searching, please refer to following Fig. 4 and related description thereof on a large scale;
Step 308: carry out one and search among a small circle, it is a trickle seek operations (fine search), is used for more accurately determining search result.About this detailed description of searching among a small circle, please refer to following Fig. 5 and related description thereof;
Step 310: finish the circulation of displacement adaptability and search.
The relatively representative of being done in step 304 can be finished the seek operations of final mean annual increment movement vector in the situation of not sacrificing arithmetic speed if the diversity factor of the motion vector that is calculated less than this second critical value, then only needs to search among a small circle.If not, then should utilize earlier and search on a large scale, finish seek operations to search among a small circle again with the motion vector of fast speeds near search result.
See also Fig. 4, Fig. 4 is according to the disclosed operational flowchart of searching on a large scale of one embodiment of the invention.In the present embodiment, this flow process is corresponding to the step 306 among Fig. 3, and includes following steps:
Step 400: beginning is searched on a large scale;
Step 402: calculate each element motion vector difference degree in different directions in initial search S set.Be the motion vector difference degree that calculates in reaching on the directions X on the Y direction in the present embodiment, and as previously mentioned, difference degree herein can use similar in appearance to the variance (variance) shown in equation (3), (4) and represent;
Step 404: determine main motion vector difference direction.Be to decide main motion vector difference direction in the present embodiment with the directions X difference degree that relatively in step 402, calculated and the mode of Y direction difference degree.If the motion vector difference degree on the directions X is greater than the motion vector difference degree on the Y direction, then main motion vector difference direction promptly is made as directions X, otherwise it is made as the Y direction;
Step 406: set a parameter value d=0, and have smallest error function value person and be made as a centerablock with (when carrying out for the first time, being initial and searching in the middle of the S set) in the block that is searched;
Step 408: with this centerablock is central point, on this main motion vector difference direction that step 404 determined a plurality of blocks is carried out seek operations, and calculates its error function value.Please note, be that a plurality of blocks on the straight line on this main direction are searched in the present embodiment, this seek operations on a straight line can be carried out continuous a plurality of blocks, also can a plurality of blocks with fixed intervals be carried out, and decides on practical application.And this seek operations also can be carried out a plurality of blocks on same straight line not;
Step 410: if the error function value reckling of the block that is searched in step 408 is this centerablock, then execution in step 418, otherwise execution in step 412;
Step 412: set a parameter value d=0, and be made as this centerablock with having smallest error function value person in the block that is searched;
Step 414: with this centerablock is central point, (that is its vertical direction on the other direction of this main motion vector difference direction that step 404 determined, orthogonal direction) a plurality of blocks is carried out seek operations, and calculate its error function value.Please note, be that a plurality of blocks on the straight line on this vertical direction are searched in the present embodiment, this seek operations in a straight line can be carried out continuous a plurality of blocks, also can a plurality of blocks with fixed intervals be carried out, and decides on practical application.And this seek operations also can be carried out a plurality of blocks on same straight line not;
Step 416: if the error function value reckling of the block that is searched in step 414 is this centerablock, then execution in step 422, otherwise execution in step 406;
Step 418: if parameter value d=1, then execution in step 426, otherwise execution in step 420;
Step 420: setup parameter value d=1, get back to step 414;
Step 422: if parameter value d=1, then execution in step 426, otherwise execution in step 424;
Step 424: setup parameter value d=1, get back to step 408;
Step 426: finish to search on a large scale.
The operation of above-mentioned step 406～424, be after step 404 has determined a main motion vector difference direction, at first in the enterprising line search operation of this main motion vector difference direction, block with error function value minimum in the middle of finding out, be the center with this block again, on another direction, carry out same seek operations.The seek operations of above-mentioned two directions can continue to carry out alternately, and error function value reckling is till being set central point on the both direction in the block that is found.
See also Fig. 5, Fig. 5 is according to the disclosed operational flowchart of searching among a small circle of one embodiment of the invention.In the present embodiment, this flow process is corresponding to the step 308 among Fig. 3, and includes following steps:
Step 500: beginning is searched among a small circle;
Step 502: to a centerablock with and on every side adjacent block carry out seek operations and error of calculation function.Please note this one among a small circle seek operations can search this centerablock and adjacent all around or contiguous a plurality of blocks thereof, it can be searched according to the melee seek mode of knowing (small diamond searchpattern) or bitellos seek mode (large diamond search pattern) or other pattern.Detailed description about melee seek mode and bitellos seek mode, can consult the technical report ISO/IEC DTR 14496-7 of International Standards Organization (International Organisation for Standardisation), " Information technology Coding ofaudio-visual objects Part 7:Optimised reference software for codingof audio-visual objects ";
Step 504: compare the error function value of being calculated in step 502, if reckling is this centerablock in the described error function value, then execution in step 508, otherwise execution in step 506;
Step 506: will be made as new centerablock corresponding to the block of aforementioned smallest error function value, repeated execution of steps 502 again;
Step 508: finish to search among a small circle.
Please consult Fig. 6 and Fig. 7 simultaneously, Fig. 6 carries out a schematic diagram of searching on a large scale according to the displacement predictor method of one embodiment of the invention, and Fig. 7 carries out a schematic diagram of searching among a small circle according to the displacement predictor method of one embodiment of the invention.In Fig. 6, suppose the centerablock when block 602 begins for searching on a large scale, and determine that in step 404 main motion vector difference direction is after the A direction, then in step 408 to being a plurality of blocks 602 at center with centerablock 602 on the A direction, 604,606 search, if find that in step 410 error function value reckling is a block 604, then block 604 can be made as new centerablock in the step 412, and in step 414 and following step to being a plurality of blocks 604 at center with new centerablock 604 on the B direction, 608,612 search, and repeat above-mentioned steps till condition satisfies.In the present embodiment, the distance D between a plurality of blocks on the same direction is 5 pixels.
In Fig. 7, then suppose the centerablock when block 712 begins for searching among a small circle, then in step 502, centerablock 712 and adjacent block 834,836,838,840 thereof are searched, upgrade centerablock and repeat above-mentioned steps till condition satisfies according to certain pattern (person shown in Fig. 7 is the melee seek mode).
Please note, above-mentionedly only be an embodiment of displacement predictor method of the present invention, in fact, displacement predictor method of the present invention does not limit the length of predetermined space D, the pattern of searching among a small circle and the spatial distribution of direction A, B, for instance, the oblique line that direction A, B can corresponding mutually orthogonals.
Displacement predictor method in the foregoing description is to judge whether earlier to need to carry out to search and search among a small circle the motion vector of obtaining a pending block on a large scale, if can't directly set the motion vector of this pending block according to the known displacement vector, then earlier by searching convergence apace to be similar to the zone of this pending block on a large scale, then accurately lock the search block that is similar to this pending block by searching among a small circle again, just can concern the motion vector that calculates this pending block at last according to the spatial distribution of the search block of choosing and this pending block.In sum, above-mentioned displacement predictor method can significantly reduce search process to realize finding fast the purpose of motion vector, therefore can further promote the efficient of video files compression.
The above only is preferred embodiment of the present invention, and all equalizations of doing according to claim of the present invention change and modify, and all should belong to the covering scope of patent of the present invention.
Claims (9)
1. displacement predictor method, with deciding a motion vector as a result, this displacement predictor method includes:
Since initial displacement vector, carry out one first seek operations, with motion vector in the middle of determining, wherein this first seek operations is to search principle according to one first to search; And
From this centre motion vector, carry out one second seek operations, to determine this motion vector as a result, wherein this second seek operations is to search principle according to one second to search,
Wherein this first search principle includes the initial search direction that direction is searched in one of selection as this first seek operations in two search directions, wherein these two to search direction vertical mutually or be a fixed angle, and, search the enterprising line search of direction at these two in turn from this initial direction of searching.
2. displacement predictor method as claimed in claim 1, wherein this first seek operations is a rough seek operations, and this second seek operations is a trickle seek operations.
3. displacement predictor method as claimed in claim 1, wherein this second is searched principle and includes according to a preassigned pattern and search.
4. displacement predictor method as claimed in claim 3, wherein this preassigned pattern is the melee seek mode.
5. displacement predictor method as claimed in claim 3, wherein this preassigned pattern is the bitellos seek mode.
6. displacement predictor method, with deciding one first motion vector, this displacement predictor method includes:
Select one to search direction as an initial direction of searching in two search directions, wherein these two search directions are vertical mutually;
These initial a plurality of blocks of searching on the direction are carried out seek operations, determine one first block with an error function;
A plurality of blocks on the other direction of this initial search direction are carried out seek operations, determine one second block with this error function or another error function; And
When this first block and this second block are same block, will be made as this first motion vector corresponding to the motion vector of this block.
7. displacement predictor method as claimed in claim 6, wherein said seek operations includes the block of being searched at each, calculates an error function value.
8. displacement predictor method as claimed in claim 7, wherein this error function value is an absolute error total value, wherein this first block is this error function value reckling in the middle of these initial described a plurality of blocks of searching on direction, and this second block be central this error function value reckling of described a plurality of blocks on this initial other direction of searching direction.
9. displacement predictor method as claimed in claim 6 wherein is to decide this initial direction of searching according to the variance between a plurality of motion vectors in this selection step.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CNB2004100562645A CN100385955C (en) | 2004-08-06 | 2004-08-06 | Block displacement prognosticating method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CNB2004100562645A CN100385955C (en) | 2004-08-06 | 2004-08-06 | Block displacement prognosticating method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN1731858A CN1731858A (en) | 2006-02-08 |
| CN100385955C true CN100385955C (en) | 2008-04-30 |
Family
ID=35964140
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CNB2004100562645A Active CN100385955C (en) | 2004-08-06 | 2004-08-06 | Block displacement prognosticating method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN100385955C (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| TWI493977B (en) * | 2011-09-20 | 2015-07-21 | Hannstar Display Corp | Image searching module and method thereof |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US6130912A (en) * | 1998-06-09 | 2000-10-10 | Sony Electronics, Inc. | Hierarchical motion estimation process and system using block-matching and integral projection |
| CN1333634A (en) * | 2001-01-12 | 2002-01-30 | 北京航空航天大学 | Quick video motion estimating method |
| CN1492687A (en) * | 2003-09-04 | 2004-04-28 | 上海大学 | Quick full pixel movement evaluation method of multiple kinds of modules |
-
2004
- 2004-08-06 CN CNB2004100562645A patent/CN100385955C/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US6130912A (en) * | 1998-06-09 | 2000-10-10 | Sony Electronics, Inc. | Hierarchical motion estimation process and system using block-matching and integral projection |
| CN1333634A (en) * | 2001-01-12 | 2002-01-30 | 北京航空航天大学 | Quick video motion estimating method |
| CN1492687A (en) * | 2003-09-04 | 2004-04-28 | 上海大学 | Quick full pixel movement evaluation method of multiple kinds of modules |
Also Published As
| Publication number | Publication date |
|---|---|
| CN1731858A (en) | 2006-02-08 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US11546627B2 (en) | Moving picture decoding method and moving picture encoding method | |
| KR0171146B1 (en) | Feature point based motion vectors detecting apparatus | |
| RU2659733C2 (en) | Method and device for coding and decoding the motion vector based on the reduced predictors-candidates of the motion vector | |
| US8705611B2 (en) | Image prediction encoding device, image prediction encoding method, image prediction encoding program, image prediction decoding device, image prediction decoding method, and image prediction decoding program | |
| US7869518B2 (en) | Fast motion estimation apparatus and method using block matching algorithm | |
| KR100492127B1 (en) | Apparatus and method of adaptive motion estimation | |
| US5751362A (en) | Apparatus for encoding a video signal using feature point based motion estimation | |
| US20060008008A1 (en) | Method of multi-resolution based motion estimation and recording medium storing program to implement the method | |
| US7953153B2 (en) | Motion estimation method utilizing modified rhombus pattern search for a succession of frames in digital coding system | |
| CN101326550A (en) | Motion estimation using prediction guided decimated search | |
| Yaakob et al. | A comparison of different block matching algorithms for motion estimation | |
| KR100843418B1 (en) | Apparatus and method for image coding | |
| KR20070088322A (en) | Apparatus and method for computing motion vector | |
| KR100994768B1 (en) | Motion estimation method for encoding motion image, and recording medium storing a program to implement thereof | |
| Lin et al. | Fast intra coding unit partition decision in H. 266/FVC based on spatial features | |
| KR20050097936A (en) | Efficient predictive image parameter estimation | |
| US6496539B2 (en) | Method and apparatus for detecting motion | |
| JP4669517B2 (en) | How to realize motion estimation | |
| CN106954065B (en) | Recursive prediction image compression method based on gradient direction histogram | |
| US20050105620A1 (en) | Motion vector detecting device and motion vector detecting program | |
| JP2010521118A (en) | Multiframe video estimation from compressed video sources | |
| CN100385955C (en) | Block displacement prognosticating method | |
| US6020925A (en) | Method and apparatus for encoding a video signal using pixel-by-pixel motion prediction | |
| US9215474B2 (en) | Block-based motion estimation method | |
| US9225994B2 (en) | Global motion estimation using reduced frame lines |
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 |