CN104244007A

CN104244007A - Image compression method and device based on arbitrary shape matching

Info

Publication number: CN104244007A
Application number: CN201410259514.9A
Authority: CN
Inventors: 林涛
Original assignee: Shanghai Tianhe Electronic Information Co Ltd
Current assignee: Shanghai Tianhe Electronic Information Co Ltd
Priority date: 2013-06-13
Filing date: 2014-06-12
Publication date: 2014-12-24
Anticipated expiration: 2034-06-12
Also published as: CN111800640B; CN111800640A; CN111800641B; CN111800641A; CN104244007B

Abstract

The invention relates to an image compression method and device based on arbitrary shape matching. The image compression method and device based on arbitrary shape matching are basically characterized in that when the current coding unit performs coding, search is performed in a set of coded historical pixels (reconstruction pixels) according one or more preset matching patterns, and multiple optimal matching strings (blocks) matched with pixel samples of the current coding unit performing coding are found one by one. Each matching string (block) can have any shape and can be represented with two basic matching parameters, namely the matching position and the matching size.

Description

Based on method for compressing image and the device of arbitrary shape coupling

Technical field

The present invention relates to a kind of compression of digital video coding and decode system, particularly comprise the digital complex image of continuous-tone image and non-continuous tone image and the coding of video and the method and apparatus of decoding.

Background technology

Along with cloud-mobile computing, intelligent TV network (IPTV), Wireless Display connect the development of (WIFI Display etc.) with universal, interconnected between multicomputer, between main frame and other digital devices (as intelligent television, smart mobile phone, panel computer etc.) and between various all kinds of digital device has become a reality and has more and more become a kind of main trend.This makes server end (" cloud " or " cloud sheet ") become current active demand to the real time screen transmission of user side.Owing to needing the screen video data volume of transmission very large, for panel computer 2048x1536 pixel resolution and 60 frames/second refresh rate 24 true color screen pictures, the data of transmission are needed to reach 2048x1536x60x24=4320 megabit per second, it is impossible that so many data want under current network condition, realize real-time Transmission, and therefore effective data compression is essential.

Take computer screen image as the digital complex image of representative and the outstanding feature of video be comprise word, figure, line segment and natural image.Wherein, word, figure, line segment, generally by Practical computer teaching, have the characteristic of non-continuous tone, and natural image is obtained by video camera shooting, has the characteristic of continuous tone.Current, to the compressed encoding comparative maturity of natural image and video, there are many set standards to adopt.The framework of the hybrid coding that these standards adopt predictive coding to be combined with transition coding and quantization encoding substantially, but they are poor to the compression effectiveness of screen picture, must seek to be more suitable for the algorithm of screen picture compression to realize effective compression.

Making full use of the feature of computer screen image, computer screen image is carried out to the compression of Ultra-High Efficiency, is also the up-to-date international video compression standards HEVC(High Efficiency Video Coding formulated) a main target.

The natural form of the digital video signal of screen picture is the sequence of image.The rectangular area that piece image is normally made up of some pixels, if digital video signal is per second 50 width images, so the digital video signal of 30 minutes is exactly a sequence of video images be made up of 30x60x50=90000 width image, sometimes also referred to as video sequence or sequence.Digital encoding video signal is encoded to a width piece image according to a definite sequence exactly.At any one time, that piece image of encoding is called current encoded image.Equally, carrying out decoding to the compressed bit stream of digital video signal is exactly decode to the compressed bit stream of a width piece image according to same order.At any one time, that piece image of decoding is called current decoded picture.Current encoded image or current decoded picture are all referred to as present image.

In the international standard of nearly all encoding video pictures as MPEG-1/2/4, H.264/AVC and in HEVC, when piece image is encoded, it is all subgraph piece image being divided into some pieces of MxM pixels, be called " coding unit (Coding Unit is called for short CU) ", take CU as basic coding unit, subgraph one piece a piece is encoded.The size of conventional M is 8,16,32,64.Therefore, carry out coding to a sequence of video images to encode successively to each coding unit exactly.Equally, be also that each coding unit is decoded successively during decoding, finally reconstruct whole sequence of video images.

For adapting to the different of each several part picture material and character in piece image, carry out most efficient coding targetedly, in piece image, the size of each CU can be different, has plenty of 8x8, has plenty of 64x64, etc.In order to enable the CU of different size seamless spliced, piece image is always first divided into that size is identical to be had nx n" maximum coding unit (Largest Coding Unit is called for short LCU) " of pixel, then each LCU Further Division becomes 1 or 4 or 16 or 64 identical CU of size.Such as, piece image is first divided into the LCU(of the identical 64x64 pixel of size n=64).Then, the wherein CU of some LCU inherently 64x64 pixel, be namely called that the degree of depth is the CU of 1 to the CU(of 1 64x64 pixel, a CU has nx nindividual pixel, it is the division of 1 that this CU forms a degree of depth) encode.Some LCU is then divided into the CU of 4 32x32 pixels, and be namely called that the degree of depth is the CU of 2 to the CU(of 4 32x32 pixels, each CU has x individual pixel, it is the division of 2 that these 4 CU form a degree of depth) encode.Also have some LCU to be then divided into the CU of 16 16x16 pixels, be namely called that the degree of depth is the CU of 3 to the CU(of 16 16x16 pixels, each CU has x individual pixel, it is the division of 3 that these 16 CU form a degree of depth) encode.Other LCU is then divided into the CU of 64 8x8 pixels, and be namely called that the degree of depth is the CU of 4 to the CU(of 64 8x8 pixels, each CU has x individual pixel, it is the division of 4 that these 64 CU form a degree of depth) encode.

When encoding to a LCU, first the division that the degree of depth is 1 is encoded, secondly the division (i.e. these all CU divided) that the degree of depth is 2 is encoded, so successively the division of different depth is encoded, until the degree of depth reaches a depth capacity D preset, namely the size of corresponding CU reaches minimum value x till.After completing the whole codings to the division of D different depth, from the coding of the division of these different depths, according to rate-distortion performance, the division of the MxM pixel CU of an optimum degree of depth and a kind of optimum is selected to carry out final coding to this LCU.

Each LCU also can be divided into the phasing multiple not of uniform size CU together of tree in another way.Therefore, LCU is also referred to as " code tree unit (Coding Tree Unit is called for short CTU) ".In patent application of the present invention, LCU and CTU is synonym.The CU onesize with CTU is called that degree of depth D is 0(D=0) CU.The CU of 1/4th sizes that the CU quartering of a D=0 is obtained calls that degree of depth D is 1(D=1) CU.The CU of the D=1 CU that the size that obtains of the quartering is less is again called that degree of depth D is 2(D=2) CU.The CU of the D=2 CU that the size that obtains of the quartering is less is again called that degree of depth D is 3(D=3) CU.The CU of the D=3 CU that the size that obtains of the quartering is less is again called that degree of depth D is 4(D=4) CU.Such as, piece image is first divided into the LCU(N=64 of the identical 64x64 pixel of size).One of them LCU can be made up of the CU of the D=3 of the CU and 8 of the D=2 of the CU of the D=1 of 2 32x32 pixels, a 6 16x16 pixels 8x8 pixel.Such 16 become the CU of tree to form a CTU.Wherein another LCU can be made up of the CU of the D=2 of the CU of the D=1 of 3 32x32 pixels and 4 16x16 pixel, and such 7 become the CU of tree also to form a CTU.

Encoding to piece image, is exactly encode to the CU one by one in CTU one by one successively.At any one time, the CU encoded is called present encoding CU.Decoding to piece image, is also decode to the CU one by one in CTU one by one successively according to same order.At any one time, the CU decoded is called current decoding CU.Present encoding CU or current decoding CU is referred to as current C U.

Pixel has various form, as: the rgb pixel form that 1 pixel is made up of 1 red component (component), 1 green component, 1 blue component; The YUV4:4:4 pixel format that 1 pixel is made up of 1 luminance Y component, 1 colourity U component, 1 colourity V component; The YUV4:2:2 pixel format that adjacent 2 pixels in left and right are made up of 2 luminance Y component, 1 colourity U component, 1 colourity V component; The YUV4:2:0 pixel format that 4 pixels of the neighbouring 2x2 of the pressing spatial arrangements in left and right are made up of 4 luminance Y component, 1 colourity U component, 1 colourity V component.One-component generally represents by the numeral of 18 ~ 16 bit.YUV4:2:2 pixel format and YUV4:2:0 pixel format are all obtain the down-sampling of YUV4:4:4 pixel format execution chromatic component.Be commonly referred to as that the color format of YUV is actual comprises multicolour form, as YCbCr color format.Most basic element when coding or decoding can be a pixel, also can be a pixel component.Be referred to as a pixel sample value (sample) as a pixel of the most basic element of coding or decoding or a pixel component, be sometimes also commonly referred to as a pixel value, or be called a sample value simply.When a CU is encoded, a CU can be divided into 3 component planes (G plane, B plane, R plane or Y plane, U plane, V plane), 3 component planes are encoded respectively; Also 3 of a pixel component binding combination can be become 3 tuples, the CU entirety be made up of these 3 tuples is encoded.The arrangement mode of front a kind of pixel and component thereof is called the planar format (planar format) of image (and CU), and then the arrangement mode of a kind of pixel and component thereof is called the stacked package form (packed format) of image (and CU).

In patent application of the present invention, " pixel sample value ", " pixel value ", " sample value " are synonyms, based on context, clearly can represent " pixel " or represent " pixel component " or both expressions simultaneously.If can not be clear and definite from context, be exactly so represent both simultaneously.

Comprise in all kinds image of screen picture and the coding techniques of video sequence, use the schematic flow sheet of the coding method of maximum prior aries as shown in Figure 1.The coding method of prior art comprises the steps:

1) reading in a size is the maximum coding unit of NxN pixel and the original pixels of LCU;

2) initial value of variable M is set to N, is namely ready to first encode to the division that size is the current coded unit CU of MxM=NxN pixel;

3) intraframe predictive coding based on neighborhood pixels sample value and the interframe based on square (block) (namely between current encoded frame and former encoded frame) predictive coding are carried out to the CU of 1 MxM pixel, be referred to as predictive coding, produce (1) prediction residual and (2) predictive mode and motion vector.Need in predictive coding to attempt multiple intra prediction mode and multiple inter-frame forecast mode (being referred to as predictive mode).Each predictive mode all can produce corresponding prediction residual and experience subsequent step 4), 5), 6), 7) computing;

4) transition coding is carried out to one of the coding result of step 3) prediction residual, produce conversion coefficient.Transition coding is optional, if namely transition coding does not reach better data compression effects, does not just carry out transition coding;

5) quantization encoding is carried out to conversion coefficient (having carried out the situation of transition coding in previous step) or prediction residual (not carrying out the situation of transition coding in previous step), produce quantization transform coefficient or quantitative prediction residual error.In the situation of lossless coding, quantization encoding computing is identity of operation, does not produce the quantification effect of any reality;

6) to step 3) ~ 5) coding result carry out the inverse operation of described coding, namely preliminary reconstruction goes out the pixel of described CU, for subsequent step 8) rate-distortion cost (Rate-distortion cost) calculate, and at subsequent step 11) determine to present encoding LCU optimal dividing after, all preliminary reconstruction pixels of optimal dividing are carried out to the reprocessing of block-eliminating effect filtering and pixel compensation computing, then the pixel of each reconstruct CU through reprocessing is put into history pixel (reconstructed pixel) working area, as the reference pixel that subsequent prediction is encoded and reconstructed.Because coding may damage, reconstructed pixel might not equal to input original pixels;

7) two predictive modes of the coding result of step 3) and the quantization encoding result of motion vector and step 5) and quantization transform coefficient and quantitative prediction residual error are carried out to entropy code and produce the bit rate of compressed bit stream.Entropy code also comprises the differential coding or the index coding that first carry out tieing up based on 1 dimension or 2 contiguous sample values to the sample value of predictive mode, motion vector, quantization transform coefficient, quantitative prediction residual error constant entropy coded object, to remove the correlation between sample value, improve the efficiency of entropy code;

8) from bit rate or the bit rate estimated value of original pixels, reconstructed pixel, compressed bit stream, calculate rate-distortion cost, select the optimal prediction modes of this CU according to rate-distortion performance and the compressed bit stream data of this CU temporary;

9) judge whether to complete the coding to all MxM pixel CU, if so, then order performs next step, otherwise skips to the coding that step 3) starts next MxM pixel CU;

10) judge whether that M has reached the minimum value preset, if so, then order performs next step, otherwise, M is reduced by half, then skips to the coding that step 3) starts all MxM pixel CU in the division of the next degree of depth;

11) from the coding result of the division of completed multiple different depth (CU of different size MxM pixel), select the division of the CU of an optimum MxM pixel according to rate-distortion performance and export the corresponding compressed bit stream data of this division;

12) judge whether the coding completing all LCU, if so, then terminate coding, otherwise get back to step 1), start the coding to next LCU.

The schematic flow sheet of the coding/decoding method of prior art as shown in Figure 2.The coding/decoding method of prior art comprises the steps:

1) Partial shrinkage bit stream data is read in;

2) carry out the entropy decoding of a coding unit and CU, obtain header and the data message of this CU.Header mainly comprise this CU adopt based on neighborhood pixels sample value infra-frame prediction or and based on the inter prediction of square, whether carry out inverse transformation decoding.Entropy decoding also comprises differential decoding or the index decoding of tieing up contiguous sample values based on 1 dimension or 2 to predictive mode, motion vector, quantization transform coefficient, quantitative prediction residual error constant entropy decoder object;

3) inverse quantization decoding is carried out to quantization transform coefficient or quantitative prediction residual error, produce conversion coefficient or prediction residual;

4) if previous step produce be conversion coefficient, then carry out inverse transformation decoding, produce prediction residual, otherwise, order perform next step;

5) carry out, based on the infra-frame prediction decoding of neighborhood pixels sample value or the interframe prediction decoding based on square, being referred to as prediction decoding, producing preliminary reconstructed pixel;

6) preliminary reconstruction pixel is carried out to the reprocessing of block-eliminating effect filtering and pixel compensation computing, then the pixel of the reconstruct CU through reprocessing is put into history pixel (reconstructed pixel) working area, as the reference pixel of subsequent prediction decoding;

7) pixel of reconstruct CU is exported;

8) judge whether to complete the decoding to the compressed bit stream data of all CU in a LCU, if so, then order performs next step, otherwise gets back to step 1), start the decoding to next CU;

9) judge whether the decoding of the compressed bit stream data completed all LCU, if so, then terminate decoding, otherwise get back to step 1), start the decoding to next LCU.

The code device schematic diagram of prior art as shown in Figure 3.Whole code device forms by with lower module:

1) memory module of the pixel of at least one coding unit and CU can be stored;

2) predictive coding module, implements based on the intraframe predictive coding of neighborhood pixels and the inter prediction encoding based on square (block) input pixel sample value, and exports (1) prediction residual and (2) predictive mode and motion vector;

3) conversion module.Transform operation is implemented to prediction residual, output transform coefficient.To the screen picture pixel of some type, transform operation can not play the effect of packed data, and in this situation, just do not implement transform operation, namely conversion module is bypassed, direct prediction of output residual error;

4) quantization modules.Quantization operations is implemented to conversion coefficient (situation in the non-bypass of conversion module) or prediction residual (situation conversion module is bypassed), produces quantization transform coefficient or quantitative prediction residual error;

5) entropy code module.Entropy code is implemented to predictive mode, motion vector, quantization transform coefficient or quantitative prediction residual error, comprises and the differential coding tieing up contiguous sample values based on 1 dimension or 2 is first implemented to the sample value of these entropy code objects;

6) reconstruct and post-processing module.Implement the inverse operation of predictive coding module, conversion module, these three modules of quantization modules, reconstruct the pixel of described CU and carry out the reprocessing of block-eliminating effect filtering and pixel compensation computing, then the pixel of the reconstruct CU through reprocessing is put into history pixel (reconstructed pixel) temporary storage module, as the predictive coding of follow-up CU and the reference pixel of reconstruct;

7) history pixel (reconstructed pixel) temporary storage module.This module provides the reference pixel of predictive coding and reconstruct;

8) based on optimal prediction modes and the division selection module of rate-distortion performance.When encoding each CU, select the optimal prediction modes of this CU according to rate-distortion performance.Complete all different depths to a LCU division all CU coding after, select the degree of depth of the optimum of this LCU according to rate-distortion performance and corresponding to divide;

9) compressed bit stream output module.The coded compressed data of the optimal dividing selected is outputted to output video compressed bit stream.

Prior art decoding device schematic diagram as shown in Figure 4.Whole decoding device forms by with lower module:

1) Partial shrinkage bit stream data memory module;

2) entropy decoder module.Entropy decoding is implemented to input compressed bit stream data.Entropy decoding also comprises the differential decoding tieing up contiguous sample values based on 1 dimension or 2 to predictive mode, motion vector, quantization transform coefficient, quantitative prediction residual error constant entropy decoder object.When entropy decoding also comprises the degree of depth, the coding of division when to parse coding from input compressed bit stream data, whether transform operation the decoding auxiliary information such as is bypassed;

3) inverse quantization module.Implement inverse quantization computing, output transform coefficient or prediction residual;

4) inverse transform module.If transform operation is not bypassed during coding, then implement inverse transformation decoding, prediction of output residual error, otherwise, do not implement inverse transformation decoding, direct prediction of output residual error;

5) prediction decoding module.Prediction residual is implemented based on the infra-frame prediction decoding of neighborhood pixels sample value or the interframe prediction decoding based on square, exports preliminary reconstructed pixel;

6) post-processing module.Preliminary reconstruction pixel is implemented to the reprocessing of block-eliminating effect filtering and pixel compensation computing, then the reconstructed pixel through reprocessing is put into history pixel (reconstructed pixel) temporary storage module, as the reference pixel of subsequent prediction decoding;

7) history pixel temporary storage module.This module provides the reference pixel of prediction decoding;

8) reconstructed pixel output module.Reconstructed pixel is the last output of decoding device.

As mentioned above, in prior art, the first step of coding is all carry out the intraframe predictive coding based on neighborhood pixels sample value or the interframe based on square (block) (namely between current encoded frame and former encoded frame) predictive coding to coding unit CU.In the occasion that entire image is all natural image, prior art is effective.

Along with multimedia technology in a computer universal, in the screen picture of today and following routine use computer, the bitmap (bitmap) usually containing the composition such as a lot of letter, numeral, word, menu, small icon, large figure, chart, form in piece image.Much identical or similar pattern (pattern) is had in the image that one width is such.As English only has 52 different letters (26 capitalizations and 26 lowercases), Chinese only has little several different stroke to form especially.Therefore, in computer screen image, find matched pixel sample value string or the matched pixel sample value block of arbitrary shape, just can go here and there distance between (block) (1 ties up apart from or 2 ties up distance) these two parameters and represent to be mated and go here and there the full detail of (block) by the size of coupling string (block) with mate, eliminate the redundancy existed in image pixel, play significant Image Data Compression effect.

But, in prior art, based on the intraframe predictive coding of neighborhood pixels sample value and the inter prediction encoding based on square (block), the pattern that the arbitrary shape in image mates all effectively can not be found, very low to the code efficiency of this kind of image and pattern.

Summary of the invention

In order to solve this problem of the prior art of image/video Code And Decode, the invention provides a kind of based on the Image Coding of arbitrary shape coupling and the method and apparatus of decoding.

Technical characteristics of the present invention as shown in Figure 5 and Figure 6.

In coding method of the present invention and device, the most basic peculiar technical characteristic is exactly when encoding to current coded unit CU, according to one or more prespecified pixel sample values arrangement scan format and coupling form (both are collectively referred to as match pattern), search in the history pixel (reconstructed pixel) set (also referred to as reference pixel sample value collection) completing coding, that finds multiple optimum matched with the pixel sample value of present encoding CU one by one mates string (block).Each coupling string (block) can have arbitrary shape and can by two basic match parameter, and namely matched position (also claiming matching distance) p and matched size s characterizes:

1) matched position p be find in the set of history pixel corresponding coupling string (block) first pixel sample value with in present encoding CU by mate go here and there (block) first pixel sample value between linear (1 ties up) apart from or plane (2 tie up) distance, namely relative position, its unit is pixel sample value.Such as, p=125 pixel sample value represents that linear range is 125 pixel sample values, and p=(34,12) pixel sample value represents that horizontal range and vertical range are 34 pixel sample values and 12 pixel sample values respectively.

2) matched size is the 1 dimension length (or 2 dimension areas) of coupling string (or block).Its unit is also pixel sample value.Such as, s=100 pixel sample value represents coupling string and is had 100 pixel sample values all respectively by coupling string, s=8x12 pixel sample value represents match block and had 8x12=96 pixel sample value all respectively by match block.Obviously, in the occasion of match block, the size s of a block generally needs to represent with multiple parametric component.Such as, block is the situation of rectangle, and s generally contains width and the height (sometimes also referred to as long and wide) of two components and rectangle.Again such as, block is the situation of parallelogram, and s generally also contains the end and the height of two components and parallelogram.Again such as, block is trapezoidal situation, and s generally contains three components and trapezoidal upper base, goes to the bottom and height.

This multiple arbitrary shape coupling string (block) adds the complete square shape of possible non-matched pixel value seamless spliced just in time formation present encoding CU.Fig. 5 has the example of the arbitrary shape coupling of 3 kinds of match patterns (comprising pixel sample value arrangement scan format and coupling form).These 3 CU of CU m in the piece image that Fig. 5 represents, CU m+1, CU m+2 adopt 3 kinds of match patterns to carry out arbitrary shape coupling respectively.In general, in a width present image, same match pattern can be all used by all CU, also different match patterns can be used adaptively by different CU.

Coupling string (block), also referred to as coupling reference strings (block), the size of the reference pixel sample value collection at place is a predetermined Code And Decode parameter.If this parameter does not write compressed bit stream, decoder just must be set as the current decoded picture of view picture the size of reference pixel sample value collection, and reserved enough memory spaces, probably exceed actual demand, and that strengthens decoder realizes cost.Therefore, the size write compressed bit stream of the reference pixel sample value collection that encoder should use reality sends decoder to, makes decoder only can reserve the memory space of the reference pixel sample value collection of appropriate size, reaches the highest operating efficiency.

What reference pixel sample value collection stored is in present image through the reconstructed pixel value of coding or decoding, and reconstructed pixel value just reaches after normally living through multiple working procedure and reconstructs completely.Therefore, the reconstructed pixel value stored in reference pixel sample value collection generally comprises the reconstructed pixel that namely different types of Perfect Reconstruction pixel and partial reconfiguration pixel in various degree only live through process portion.Different match patterns can adopt the reconstructed pixel of one species as with reference to pixel, and different types of reconstructed pixel also can be adopted as reference pixel.In same match pattern, different types of reconstructed pixel also can be adopted as reference pixel at different matched positions.

example 1: the match pattern adopting vertical scanning 1 to tie up string shape coupling carries out arbitrary shape coupling

The sequence number of Fig. 5 is the CU(CU m of m) adopt the match pattern of vertical scanning 1 dimension string shape coupling to carry out arbitrary shape coupling.Tie up in the match pattern of string shape coupling at vertical scanning 1, history pixel is first with the arrangement of the order of LCU sequence number, and in a LCU, pixel sample value arranges by column with vertical scanning form, and like this, history pixel is finally arranged in the pixel sample value string of an one dimension.Such as, in Figure 5, the size of a LCU is 16x16 pixel, and each pixel has 3 sample values (component), and the degree of depth is 1, and therefore the size of a CU is also 16x16 pixel.First pixel sample value in the one-dimensional pixel sample value string that history pixel is arranged in be sequence number be 0 LCU and CU(CU 0) first, the top pixel sample value of the 1st row.In one row pixel, arrange from top to bottom with vertical scanning form.Therefore, second pixel sample value in described one-dimensional pixel sample value string is that the top number of the 1st row of CU 0 gets off second pixel sample value.After 1st row, arrangement is the 2nd row, is equally also arrange pixel sample value from top to bottom with vertical scanning form.So by column arrangement until CU 0 the 16th row, what arrange afterwards is that in Fig. 5, sequence number is LCU and CU(CU 1 of 1) first row pixel, the like.In piece image shown in Fig. 5, horizontal direction has h LCU and CU.Therefore, sequence number is LCU and the CU(CU h-1 of h-1) the 16th row pixel be last (the rightest) 1 row pixel of first LCU capable (total h LCU and CU), LCU and the CU(CU h of to be sequence number the be h arranged afterwards) the 1st row pixel, the most left 1 row pixel that namely second LCU is capable.The sequence number of Fig. 5 is that first 3 of depicting in LCU and the CU m of m in arbitrary shape coupling are gone here and there by coupling:

1) first is had 42 pixel sample values (=14 pixels) by coupling string (being numbered the pixel sample value string of in circle 1 in the CU m of Fig. 5).The coupling string of the correspondence found in history pixel sequence number be 0 LCU and CU 0 and sequence number be (numbering be also in circle 1) in LCU and CU 1 of 1, front 5 pixels (15 pixel sample values) are last 5 pixels of the 16th row of CU 0, and then 9 pixels (27 pixel sample values) are the most front 9 pixels of the 1st row of CU 1.First pixel sample value of coupling string is the 34th pixel sample value of the 16th row of CU 0, so linear (1 dimension) address of its position is

Pixel sample value numbering=768 x 0+48 x 15+34=754 in pixel sample number x (column number-the 1)+row of the pixel sample number xLCU sequence number+row of a LCU.

And by the 1st pixel sample value of the 1st row of first pixel sample value of coupling string to be sequence number be LCU and CU m of m, so linear (1 dimension) address of its position is

768 x m＋48 x 0＋1＝768m＋1。

So this matched position being numbered the arbitrary shape coupling of in circle 1 is p=768m+1-754=768m-753.And its matched size (length) is obviously s=42.The unit of p and s is all pixel sample value.

2) second is had 33 pixel sample values (=11 pixels) by coupling string (being numbered the pixel sample value string of in circle 2 in the CU m of Fig. 5).The coupling string of the correspondence found in history pixel in CU 0 (numbering be also in circle 2), front 7 pixels (21 pixel sample values) are last 7 pixels of the 3rd row of CU 0, and then 4 pixels (12 pixel sample values) are the most front 4 pixels of the 4th row of CU 0.First pixel sample value of this coupling string is the 28th pixel sample value of the 3rd row of CU 0, and mated the 43rd the pixel sample value that first pixel sample value of going here and there is the 1st row of CU m, so matched position that this arbitrary shape being numbered in circle 2 mates is p=(768m+48 x 0+43)-(768 x 0+48 x 2+28)=768m-81.And its matched size (length) is obviously s=33.

3) the 3rd is had 45 pixel sample values (=15 pixels) by coupling string (being numbered the pixel sample value string of in circle 3 in the CU m of Fig. 5).The coupling string of the correspondence found in history pixel in CU h-1 and CU h (numbering be also in circle 3), front 13 pixels (39 pixel sample values) are last 13 pixels of the 16th row of CU h-1, and then 2 pixels (6 pixel sample values) are the most front 2 pixels of the 1st row of CU h.First pixel sample value of this coupling string is the 10th pixel sample value of the 16th row of CU h-1, and mated the 28th the pixel sample value that first pixel sample value of going here and there is the 2nd row of CU m, so matched position that this arbitrary shape being numbered in circle 3 mates is p=(768m+48 x 1+28)-(768 x (h-1)+48 x 15+10)=768 (m-h+1)-654.And its matched size (length) is obviously s=45.

example 2: adopt the match pattern that vertical scanning 2 repair and maintenance shape (2D-shape-preserved) mates to carry out arbitrary shape coupling

The sequence number of Fig. 5 is the present encoding CU(CU m+1 of m+1) adopt the match pattern that vertical scanning 2 repair and maintenance shape is mated to carry out arbitrary shape coupling.In the match pattern of vertical scanning 2 repair and maintenance shape coupling, history pixel retains 2 dimension array formats of original image, and in one pixel, 3 pixel component (sample value) are vertical arrangements, and in present encoding CU, pixel sample value then arranges by column with vertical scanning form and arranges from top to bottom in 1 row.When searching for matched pixel sample value string in history pixel, in present encoding CU, moved from top to bottom with vertical scanning form by matched pixel sample value, after a column scan and coupling are over, then scan and mate the adjacent row in the right.The matched pixel sample value string found in history pixel must with kept on all four 2 to tie up shape by matched pixel sample value string in present encoding CU.Depict in the CU m+1 of Fig. 5 and adopt first 2 in the arbitrary shape coupling of vertical scanning 2 repair and maintenance shape match pattern to be gone here and there by coupling:

1) first is had 45 pixel sample values (=15 pixels) by coupling string (being numbered the pixel sample value string of in circle 4 in the CU m+1 of Fig. 5).The coupling string of the correspondence found in history pixel in CU 1 and CU h+1 (numbering be also in circle 4).The border of this coupling string leap 2 CU, 3 pixels (9 pixel sample values) are in CU 1, and other 12 pixels (36 pixel sample values) are in CU h+1.Coupling string in the set of history pixel has identical 2 tie up shape with the string that mated in present encoding CU, and namely, coupling string is 45 pixel sample values that formation 1 arranges, and is 45 pixel sample values that formation 1 arranges too by coupling string.First pixel sample value of coupling string is the 40th pixel sample value of the 3rd row of CU 1, so the plane of its position (2 dimension) coordinate address is

Pixel columns=18 of horizontal coordinate=this left, position,

Pixel sample number=39 above vertical coordinate=this position.

And mated the 1st the pixel sample value that first pixel sample value of going here and there is the 1st row of CU m+1.If the left of present encoding CU m+1 has the h row pixel of this two field picture and top to have v the pixel sample value of this two field picture, by the plane of the position of first pixel sample value of coupling string (2 dimension) coordinate address be then

Horizontal coordinate=h+0=h,

Vertical coordinate=v+0=v.

So this matched position being numbered arbitrary shape coupling of in circle 4 is p=(h-18, v-39), also i.e. coupling string and be h-18 by the horizontal range between two corresponding pixel sample values any in coupling string and vertical range is v-39.This matched size being numbered the arbitrary shape coupling of in circle 4 is obviously s=45.The unit of p and s is all pixel sample value.

2) second is had 60 pixel sample values (=20 pixels) by coupling string (being numbered the pixel sample value string of in circle 5 in the CU m+1 of Fig. 5).The coupling string of the correspondence found in history pixel spans CU 1, CU 2, CU h+1, these 4 CU of CU h+2.2 pixels (6 pixel sample values) of this coupling string (numbering is also circle interior 5) are in CU 1,2 pixels (6 pixel sample values) are in CU 2,15 pixels (45 pixel sample values) are in CU h+1, and 1 pixel (3 pixel sample values) is in CU h+2.Coupling string in the set of history pixel has identical 2 tie up shape with the string that mated in present encoding CU.Namely, coupling string is made up of 3 row, 1st shows 3 pixel sample values, 2nd shows 48 pixel sample values, 3rd shows 9 pixel sample values, and the 1st row align with the 2nd bottom arranged, and the 2nd row align with the 3rd top arranged, the vertical height (containing upper and lower end points pixel) of this coupling string is 16 pixels, is equal to the height of present encoding CU m+1.By coupling string also completely same by 3 row form, 1st shows 3 pixel sample values, 2nd shows 48 pixel sample values, 3rd shows 9 pixel sample values, 1st row align with the 2nd bottom arranged, and the 2nd row align with the 3rd top arranged, this is mated height i.e. 16 pixels that the vertical height of string (containing upper and lower end points pixel) is also present encoding CU m+1 certainly.First pixel sample value of this coupling string is the 40th pixel sample value of the 15th row of CU h+1, there are the 30 row pixels of this two field picture its left, there are 87 pixel sample values of this two field picture top, and mated the 46th the pixel sample value that first pixel sample value of going here and there is the 1st row of CU m+1, there is the h+0 row pixel of this two field picture its left, there is v+45 the pixel sample value of this two field picture top, so this matched position being numbered the arbitrary shape coupling of in circle 5 is p=(h+0-30, v+45-87)=(h-30, v-42), also i.e. coupling string and be h-30 by the horizontal range between two corresponding pixel sample values any in coupling string and vertical range is v-42.This matched size being numbered the arbitrary shape coupling of in circle 5 is obviously s=60.

In this example, the feature of match pattern is matched pixel sample value string and all formed 2 in present image by matched pixel sample value string and tie up region, it is a part for present image, these two 2 dimension regions, one in history pixel, another is in current C U, two 2 dimension regions have equal size, and keep on all four 2 dimension shapes.The row that each 2 dimension regions are formed by pixel value vertical arrangement formed, centre except first row and last row respectively arranges identical height, the i.e. height in this region, first row aligns with the bottom in this region, and last row align with the top in this region.The height in region can be equal with the height of current C U, also can not wait with the height of current C U and get the value that another one is worth arbitrarily or derives from the height of current C U, as shown in example 6 below and example 7 current C U height 1/2nd or 1/4th.First row and last row also can respectively show identical height with centre, and namely equal the height in this region, at this moment, this region is exactly a rectangle or square, as shown in example 6 below and example 7; Matching area in this example is 2 dimension regions, matched size generally needs to represent with two or more parametric component, the special case that can derive from other parameters cataloged procedure and decode procedure or variable at the height of matching area (as equal the height of current C U or its 1/2nd), matched size also can only represent with a parametric component.

example 3: adopt the match pattern that horizontal sweep 2 repair and maintenance shape (2D-shape-preserved) is mated to carry out arbitrary shape coupling

The sequence number of Fig. 5 is the present encoding CU(CU m+2 of m+2) adopt the match pattern that horizontal sweep 2 repair and maintenance shape is mated to carry out arbitrary shape coupling.In the match pattern of horizontal sweep 2 repair and maintenance shape coupling, history pixel retains 2 dimension array formats of original image, and in one pixel, 3 pixel component (sample value) are also horizontal, and in present encoding CU, pixel sample value then arranges line by line with horizontal sweep form and arranges from left to right in 1 row.When searching for matched pixel sample value string in history pixel, in present encoding CU, moved from left to right with horizontal sweep form by matched pixel sample value, after a line scanning and coupling are over, then scan and mate a line adjacent below.The matched pixel sample value string found in history pixel must with kept on all four 2 to tie up shape by matched pixel sample value string in present encoding CU.Depict in the CU m+2 of Fig. 5 and adopt first 3 in the arbitrary shape coupling of horizontal sweep 2 repair and maintenance shape match pattern to be gone here and there by coupling:

1) first is had 72 pixel sample values (=24 pixels) by coupling string (being numbered the pixel sample value string of in circle 6 in the CU m+2 of Fig. 5).The coupling string of the correspondence found in history pixel in CU 1 and CU 2 (numbering be also in circle 6).The border of this coupling string leap 2 CU, 14 pixels (42 pixel sample values) are in CU 1, and other 10 pixels (30 pixel sample values) are in CU 2.Coupling string in the set of history pixel has identical 2 tie up shape with the string that mated in present encoding CU.Namely, coupling string is made up of 2 row, and the 1st row has 48 pixel sample values, 2nd row has 24 pixel sample values, the left justify of the 1st row and the 2nd row, the horizontal width (containing left and right end points pixel) of this coupling string is 16 pixels, is equal to the width of present encoding CU m+2.Be made up of 2 row too by coupling string, 1st row has 48 pixel sample values, 2nd row has 24 pixel sample values, the left justify of the 1st row and the 2nd row, and this is mated width i.e. 16 pixels that the horizontal width of string (containing left and right end points pixel) is also present encoding CU m+2 certainly.First pixel sample value of coupling string is the 28th pixel sample value of the 2nd row of CU 1, so the plane of its position (2 dimension) coordinate address is

Pixel sample number=75 of horizontal coordinate=this left, position,

Number of lines of pixels=1 above vertical coordinate=this position.

And mated the 1st the pixel sample value that first pixel sample value of going here and there is the 1st row of CU m+2.If there is h the pixel sample value of this two field picture the left of present encoding CU m+2 and there is the capable pixel of the v of this two field picture top, by the plane of the position of first pixel sample value of coupling string (2 dimension) coordinate address be then

Horizontal coordinate=h+0=h,

Vertical coordinate=v+0=v.

So this matched position being numbered arbitrary shape coupling of in circle 6 is p=(h-75, v-1), also i.e. coupling string and be h-75 by the horizontal range between two corresponding pixel sample values any in coupling string and vertical range is v-1.This matched size being numbered the arbitrary shape coupling of in circle 6 is obviously s=72.The unit of p and s is all pixel sample value.

2) second is had 45 pixel sample values (=15 pixels) by coupling string (being numbered the pixel sample value string of in circle 7 in the CU m+2 of Fig. 5).The coupling string of the correspondence found in history pixel in CU h and CU h+1 (numbering be also in circle 7).The border of this coupling string leap 2 CU, 6 pixels (18 pixel sample values) are in CU h, and other 9 pixels (27 pixel sample values) are in CU h+1.Coupling string in the set of history pixel has identical 2 tie up shape with the string that mated in present encoding CU.Namely, coupling string is made up of 2 row, and the 1st row has 24 pixel sample values, and the 2nd row has 21 pixel sample values, and the horizontal width (containing left and right end points pixel) of this coupling string is 16 pixels, is equal to the width of present encoding CU m+2.Be made up of 2 row too by coupling string, the 1st row has 24 pixel sample values, and the 2nd row has 21 pixel sample values, and this is mated width i.e. 16 pixels that the horizontal width of string (containing left and right end points pixel) is also present encoding CU m+2 certainly.First pixel sample value of this coupling string is the 7th pixel sample value of the 15th row of CU h+1, there are 54 pixel sample values of this two field picture its left, there are the 30 row pixels of this two field picture top, and mated the 25th the pixel sample value that first pixel sample value of going here and there is the 2nd row of CU m+2, there is h+24 the pixel sample value of this two field picture its left, there is the capable pixel of the v+1 of this two field picture top, so this matched position being numbered the arbitrary shape coupling of in circle 7 is p=(h+24-54, v+1-30)=(h-30, v-29), also i.e. coupling string and be h-30 by the horizontal range between two corresponding pixel sample values any in coupling string and vertical range is v-29.This matched size being numbered the arbitrary shape coupling of in circle 7 is obviously s=45.

3) the 3rd is had 54 pixel sample values (=18 pixels) by coupling string (being numbered the pixel sample value string of in circle 8 in the CU m+2 of Fig. 5).The coupling string of the correspondence found in history pixel in CU 1 and CU 2 (numbering be also in circle 8).The border of this coupling string leap 2 CU, 3 pixels (9 pixel sample values) are in CU 1, and other 15 pixels (45 pixel sample values) are in CU 2.Coupling string in the set of history pixel has identical 2 tie up shape with the string that mated in present encoding CU.Namely, coupling string is made up of 2 row, and the 1st row has 27 pixel sample values, and the 2nd row has 27 pixel sample values, and the horizontal width (containing left and right end points pixel) of this coupling string is 16 pixels, is equal to the width of present encoding CU m+2.Be made up of 2 row too by coupling string, the 1st row has 27 pixel sample values, and the 2nd row has 27 pixel sample values, and this is mated width i.e. 16 pixels that the horizontal width of string (containing left and right end points pixel) is also present encoding CU m+2 certainly.First pixel sample value of this coupling string is the 13rd pixel sample value of the 11st row of CU 2, there are 108 pixel sample values of this two field picture its left, there are the 10 row pixels of this two field picture top, and mated the 22nd the pixel sample value that first pixel sample value of going here and there is the 3rd row of CU m+2, there is h+21 the pixel sample value of this two field picture its left, there is the capable pixel of the v+2 of this two field picture top, so this matched position being numbered the arbitrary shape coupling of in circle 8 is p=(h+21-108, v+2-10)=(h-87, v-8), also i.e. coupling string and be h-87 by the horizontal range between two corresponding pixel sample values any in coupling string and vertical range is v-8.This matched size being numbered the arbitrary shape coupling of in circle 8 is obviously s=54.

In this example, the feature of match pattern is matched pixel sample value string and all formed 2 in present image by matched pixel sample value string and tie up region, it is a part for present image, these two 2 dimension regions, one in history pixel, another is in current C U, two 2 dimension regions have equal size, and keep on all four 2 dimension shapes.Each 2 dimension regions are made up of horizontal the formed row of pixel value, middle rows except the first row and last column has identical width, the i.e. width in this region, the first row is alignd with the right part in this region, and last column is alignd with the left part in this region.The width in region can be equal with the width of current C U, also can not wait with the width of current C U and get the value that another one is worth arbitrarily or derives from the width of current C U, as shown in example 6 below and example 7 current C U width 1/2nd or 1/4th.The first row and last column also can have identical width with middle rows, namely equal the width in this region, and at this moment, this region is exactly a rectangle or square, as shown in example 6 below and example 7; Matching area in this example is 2 dimension regions, matched size generally needs to represent with two or more parametric component, the special case that can derive from other parameters cataloged procedure and decode procedure or variable at the width of matching area (as equal the width of current C U or its 1/2nd), matched size also can only represent with a parametric component.

example 4: employing vertical scanning 2 repair and maintenance shape is mated and the match pattern of column scan direction checker carries out arbitrary shape coupling

Namely the match pattern adopting vertical scanning 2 repair and maintenance shape to mate with example 2 is substantially identical, unique difference is: when searching for matched pixel sample value string in history pixel, in present encoding CU, moved from top to bottom in the 1st row with vertical scanning form by matched pixel sample value, move from bottom to top in the 2nd row, move from top to bottom again in the 3rd row, checker scanning direction, after one column scan and coupling are over, then scan and mate the adjacent row in the right.The vertical scanning direction of the coupling string in history pixel is also go here and there synchronous checker with being mated, and to scan moving direction consistent with being mated the pixel sample value of going here and there.

example 5: employing horizontal sweep 2 repair and maintenance shape is mated and the match pattern of direction of line scan checker carries out arbitrary shape coupling

Namely the pattern adopting horizontal sweep 2 repair and maintenance shape to mate with example 3 is substantially identical, unique difference is: when searching for matched pixel sample value string in history pixel, in present encoding CU, moved from left to right in the 1st row with horizontal sweep form by matched pixel sample value, move from right to left in the 2nd row, move from left to right again in the 3rd row, checker scanning direction, after one line scanning and coupling are over, then scan and mate a line adjacent below.The horizontal scan direction of the coupling string in history pixel is also go here and there synchronous checker with being mated, and to scan moving direction consistent with being mated the pixel sample value of going here and there.

example 6: employing horizontal sweep 2 repair and maintenance shape is mated and the match pattern that a kind of match block of fixing matched size is matching unit carries out arbitrary shape coupling

Namely the match pattern adopting horizontal sweep 2 repair and maintenance shape to mate with example 3 is substantially identical, and uniquely different 2 are:

1), when searching for Optimum Matching in the set of history pixel, not in units of the pixel sample value string of random length, but in units of the block of pixels of predetermined a kind of fixing matched size.As, in units of the square of 4x4 pixel, or the rectangle of 8x4 pixel is unit, or the rectangle of 4x8 pixel is unit.It is all rectangle or square that all namely in example 2 and example 3 arrange by pixel value the 2 dimension regions formed.

2) due in a present encoding CU, the matched size of all match block is all a predetermined fixed dimension, so any one match block in this CU, can characterize by this parameter of matched position p, and not need to characterize by matched position p and matched size s two parameters.

example 7: employing horizontal sweep 2 repair and maintenance shape is mated and the match pattern that the match block of several fixing matched size is matching unit carries out arbitrary shape coupling

Substantially identical with example 6, uniquely different 2 are:

1), when searching for Optimum Matching in the set of history pixel, not in units of the block of pixels of predetermined a kind of fixing matched size, but in units of the block of pixels of predetermined several fixing matched size.As, in units of the square of the predetermined matched size of 4x4 pixel, 8x4 pixel, 4x8 pixel these 3 kinds and rectangle.It is all several predetermined rectangle or square that all namely in example 2 and example 3 arrange by pixel value the 2 dimension regions formed.

2) to this several fixing matched size, this predetermined several matched size can be represented by numbering.So any one match block in this CU, can characterize by these two parameters of numbering of matched position p and predetermined several matched size, and not need to characterize by matched position p and matched size s two parameters; The numbering of matched size takies less bit number than matched size itself usually.

In coding/decoding method of the present invention and device, the most basic peculiar technical characteristic is exactly when the compressed bit stream decoding data to current decoding CU, first from bit stream data, parse the match pattern (example of the match pattern as shown in above example 1 ~ example 7) of the arbitrary shape coupling that this CU adopts, from bit stream data, then read the characterization parameter mating string (block) one by one and matched position and matched size successively.At some match pattern, CU or LCU only has predetermined several even a kind of fixing matched size, therefore, what each coupling string or match block read from bit stream data is not matched size itself, but the numbering of this several predetermined matched size, even just there is no the information of the respective matched size of each coupling string or match block in code stream, the information of the matched size only having this CU or LCU common at all.After the matched position obtaining coupling string (block) and matched size, decoding effort is exactly the match pattern according to arbitrary shape coupling, from the position being calculated first pixel sample value of the coupling string (block) the set of history pixel by the position of first pixel sample value of coupling string (block) and matched position of current decoding.Then, again according to the match pattern of arbitrary shape coupling, all pixel sample numbers of being specified by matched size of whole coupling string (block) just can be copied from the set of history pixel, and by mobile for whole coupling string (block) being specified its size (length or area) by matched size and paste in current decoding by the position of coupling string (block), restore whole by coupling string (block).Copy so successively, move, paste and mate string (block) one by one, finally restore all pixels of whole current decoding CU.

Fig. 6 has the decoding example of the arbitrary shape coupling of 3 kinds of match patterns (comprising pixel sample value arrangement scan format and coupling form).These 3 CU of CU m in the piece image that Fig. 6 represents, CU m+1, CU m+2 adopt these 3 kinds of match patterns to carry out arbitrary shape coupling respectively.

The match pattern that CU m in Fig. 6 have employed vertical scanning 1 dimension string shape coupling carries out arbitrary shape coupling.Therefore, the array format of history pixel is first with the arrangement of the order of LCU sequence number, and in a LCU, pixel sample value arranges by column with vertical scanning form, and like this, history pixel is finally arranged in the pixel sample value string of an one dimension.During decoding, from then on the matched position that in the set of history pixel, basis obtains from bit stream data and matched size, one by one copy number be circle in 1 coupling string, be numbered circle in 2 coupling string, be numbered circle in 3 coupling string ... Deng, and they are moved and paste current decoding CU in be numbered circle in 1 by coupling string, be numbered circle in 2 by coupling string, be numbered circle in 3 by coupling string ... Deng position, restore to be numbered circle in 1 by coupling string, be numbered circle in 2 by coupling string, be numbered circle in 3 by coupling string ... Deng.Be numbered the p=768m-753 of the coupling string of in circle 1, s=42, so be going here and there in the history pixel of shape format permutation with described 1 dimension, the place that the positional distance of going here and there is 768m-753 pixel sample value is mated from current, copy 42 pixel sample values, and these 42 pixel sample values move and paste be numbered circle in 1 by coupling string position, restore whole be numbered circle in 1 by coupling string.Equally, be numbered the p=768m-81 of the coupling string of in circle 2, s=33, so be going here and there in the history pixel of shape format permutation with described 1 dimension, the place that the positional distance of going here and there is 768m-81 pixel sample value is mated from current, copy 33 pixel sample values, and these 33 pixel sample values move and paste be numbered circle in 2 by coupling string position, restore whole be numbered circle in 2 by coupling string.With the same manner, decodable code and restore be numbered circle in 3 by coupling string ... gone here and there Deng all coupling the in CU m.

The match pattern that CU m+1 in Fig. 6 have employed vertical scanning 2 repair and maintenance shape coupling carries out arbitrary shape coupling.Therefore, history pixel retains 2 dimension array formats of original image, and in one pixel, 3 pixel component (sample value) are vertical arrangements.During decoding, from then on the matched position that in the set of history pixel, basis obtains from bit stream data and matched size, one by one copy number be circle in 4 coupling string, be numbered circle in 5 coupling string ... Deng, and they are moved and paste current decoding CU in be numbered circle in 4 by coupling string, be numbered circle in 5 by coupling string ... Deng position, restore to be numbered circle in 4 by coupling string, be numbered circle in 5 by coupling string ... Deng.Be numbered the p=(h-18 of the coupling string of in circle 4, v-39), s=45, so be tieing up in the history pixel of format permutation with 2 of image, from with current by the place of the horizontal range of mating the position of going here and there to be h-18 pixel sample value and vertical range be v-39 pixel sample value, copy 45 pixel sample values, and these 45 pixel sample values move and paste be numbered circle in 4 by coupling string position, restore whole be numbered circle in 4 by coupling string.Equally, be numbered the p=(h-30 of the coupling string of in circle 5, v-42), s=60, so be with in the history pixel of 2 of image dimension format permutation, from current by the place of the horizontal range of mating the position of going here and there to be h-30 pixel sample value and vertical range be v-42 pixel sample value, copy 60 pixel sample values, and these 60 pixel sample values move and paste be numbered circle in 5 by coupling string position, restore whole be numbered circle in 5 by coupling string.With the same manner, decodable code and all coupling of restoring in CU m+1, go here and there.

The match pattern that CU m+2 in Fig. 6 have employed horizontal sweep 2 repair and maintenance shape coupling carries out arbitrary shape coupling.Therefore, history pixel retains 2 dimension array formats of original image, and in one pixel, 3 pixel component (sample value) are also horizontal.During decoding, from then on the matched position that in the set of history pixel, basis obtains from bit stream data and matched size, one by one copy number be circle in 6 coupling string, be numbered circle in 7 coupling string, be numbered circle in 8 coupling string ... Deng, and they are moved and paste current decoding CU in be numbered circle in 6 by coupling string, be numbered circle in 7 by coupling string, be numbered circle in 8 coupling string ... Deng position, restore to be numbered circle in 6 by coupling string, be numbered circle in 7 by coupling string, be numbered circle in 8 coupling string ... Deng.Be numbered the p=(h-75 of the coupling string of in circle 6, v-1), s=72, so be tieing up in the history pixel of format permutation with 2 of image, from with current by the place of the horizontal range of mating the position of going here and there to be h-75 pixel sample value and vertical range be v-1 pixel sample value, copy 72 pixel sample values, and these 72 pixel sample values move and paste be numbered circle in 6 by coupling string position, restore whole be numbered circle in 6 by coupling string.Equally, be numbered the p=(h-30 of the coupling string of in circle 7, v-29), s=45, so be with in the history pixel of 2 of image dimension format permutation, from current by the place of the horizontal range of mating the position of going here and there to be h-30 pixel sample value and vertical range be v-29 pixel sample value, copy 45 pixel sample values, and these 45 pixel sample values move and paste be numbered circle in 7 by coupling string position, restore whole be numbered circle in 7 by coupling string.With the same manner, decodable code and restore be numbered circle in 8 by coupling string ... gone here and there Deng all coupling the in CU m+2.

By some specific instantiations, technical characteristic of the present invention is described above.Those skilled in the art can understand other advantages of the present invention and effect easily by content disclosed in the present specification.The present invention can also be implemented or be applied by embodiments different in addition, and the every details in this specification also can based on different viewpoints and application, carries out various modification or change not deviating under spirit of the present invention.

The general flow schematic diagram of coding method of the present invention as shown in Figure 7.Coding method of the present invention comprises the steps:

1) pixel characteristic analysis and evaluation, pre-treatment and coding mode selection step; Analysis and evaluation is carried out to the pixel sample value characteristic of current C U and contiguous CU, necessary pre-treatment is carried out to input pixel sample value, whether prejudge arbitrary shape coupling coded system is applicable to for encoding to current C U, this step is optional, namely can skip this step, directly enter next step; The example of the way of described analysis and evaluation: the way of foundation or the coding result with reference to contiguous CU, calculates the way of the number of the pixel of different colours in current C U; The example of described pre-treatment: sample value quantizes, color quantizing and based on color pixel cluster, carry out accurately or approx to represent to input the color of original pixels with palette and index;

2) arbitrary shape coupling coding step, arbitrary shape coupling coded system and encoded history pixel (reconstructed pixel) working area (also referred to as reference pixel sample value collection) is used to carry out arbitrary shape coupling encoding operation to current C U, the input of described arbitrary shape coupling coding is input original pixels or its pixel through pre-treatment, described arbitrary shape mates the matched position and matched size or its variant that the output of encoding is coupling string (block), non-matched pixel value or its variant, and other arbitrary shapes of coupling residual sum mate coding result parameter and variable, described matched position is used to represent and mates coupling reference pixel value that current pixel value matches the where variable parameter in described reference pixel sample value collection in described current C U, described arbitrary shape coupling coded system (is called the current string of coupling or coupling current block with the pixel sample value string of variable-size or block, its position both also can represent with a linear address with 2 dimension coordinates) encode for unit carries out coupling, the coupling that described coupling reference pixel value is formed in described reference pixel sample value collection is gone here and there or match block, be called coupling reference strings or coupling reference block, its position is also both also can represent with a linear address with 2 dimension coordinates, thus in described arbitrary shape coupling coded system, described matched position both also can use the linear address that mates reference strings (block) and the difference of linear address of mating current string (block) by 2 dimension coordinates of coupling reference strings (block) and the difference of 2 dimension coordinates mating current string (block), be commonly referred to as displacement vector to represent, because the size (equaling the size of mating current string (block)) of coupling reference strings (block) is variable, so also need another to be called, the variable parameter of matched size is to represent its size, in the occasion of coupling string, matched size is exactly generally the length of coupling string, in the occasion of match block, the size of a block generally needs to represent with multiple parametric component, such as, block is the situation of rectangle, and matched size generally contains width and the height (sometimes also referred to as long and wide) of two components and rectangle, again such as, block is the situation of parallelogram, and matched size generally also contains the end and the height of two components and parallelogram, again such as, block is trapezoidal situation, and matched size generally contains three components and trapezoidal upper base, goes to the bottom and height, described non-matched pixel value is in described reference pixel sample value collection, do not find input original pixels sample value or its pixel sample value through pre-treatment of coupling according to predetermined matching criterior, described non-matched pixel value is an input original pixels sample value or its pixel sample value through pre-treatment, so also can represent with its position in current C U, described non-matched pixel value also can be similar to by the imitative coupling sample value damaging or can't harm calculated with one, if matched size is 1, this coupling string (block) also can process as non-matched pixel value, described coupling residual error is input original pixels sample value and described difference of mating reference pixel value, if the predetermined matching criterior of arbitrary shape coupling coded system is definitely accurate harmless coupling, then described coupling residual error is zero, namely this arbitrary shape coupling coded system does not just mate residual error as output, if the predetermined matching criterior of a kind of arbitrary shape coupling coded system approximate damages coupling, then described coupling residual error may be non-vanishing, the situation that another kind damages coupling first carries out sample value quantification to input original pixels sample value, color quantizing or the pixel cluster based on color, the pre-treatment of the color of approximate representation input original pixels is carried out with palette and index, and then carry out arbitrary shape coupling coding, in this situation, because sample value quantizes, color quantizing or the pixel cluster based on color, the pixel that palette and index represent damages, even if arbitrary shape coupling coding itself is harmless, described coupling residual error (namely inputting original pixels sample value and described difference of mating reference pixel value) also may be non-vanishing, the example of described other arbitrary shapes coupling coding result: match pattern, uses coupling string to form the rule of arbitrary shape block, the type of the variant of match parameter, the computation rule of matched position, the computation rule of matched size, etc., the result of described current C U being carried out to the most fundamental sum essence of arbitrary shape coupling coding is I(I >=0) individual coupling goes here and there (block) and J(J >=0) individual non-matched pixel value, export I to (matched position, matched size) and J non-matched pixel value, described I coupling string (block) and J non-matched pixel are worth the complete square shape of seamless spliced just in time formation present encoding CU,

3) remaining matching way and various common technology are encoded and restructing operation step, complete all the other coding and restructing operations of current C U, to the input original pixels of inputted video image, various parameter and variable carry out other matching ways and various common technology, as infra-frame prediction, inter prediction, Block-matching, palette mates, bar mates, microlith mates, microlith String matching, imitative coupling sample value prediction interpolation, conversion, quantize, inverse transformation, inverse quantization, corresponding to prediction residual and the compensation (namely getting the inverse operation of residual error computing) of mating residual error, predict and ask residual error, DPCM, single order and higher difference, map, the distance of swimming, index, block-eliminating effect filtering, sample value adaptive equalization (Sample Adaptive Offset), coding and restructing operation and entropy code computing, remaining matching way described refers in coding method and is different from described step 2) matching way of the arbitrary shape that carries out coupling coding, the input of this step is above-mentioned steps 1) and step 2) output and input original pixels and the reference pixel from described reference pixel sample value collection, the output of this step is reconstructed pixel (comprising Perfect Reconstruction pixel and partial reconfiguration pixel in various degree) and the compressed bit stream containing any form fit coding result and other coding results, described reconstructed pixel is put among described reference pixel sample value collection, as follow-up arbitrary shape coupling encoding operation, remaining matching way and various common technology coding and the reference pixel required for restructing operation step, described compressed bit stream is the last output of this coding method, contains corresponding coding/decoding method and carries out decode and whole syntactic elements, particularly matched position required for reconstructing and the syntactic element such as matched size or its variant, non-matched pixel value or its variant.

The general flow schematic diagram of coding/decoding method of the present invention as shown in Figure 8.Coding/decoding method of the present invention comprises the steps:

1) compressed bit stream Data Analysis and partial decoding of h step, the matched position and matched size or its variant of encoding are mated to the arbitrary shape that contains of input, the packed data of non-matched pixel value or its variant and the compressed bit stream of every other syntactic element packed data carry out entropy decoding, and parse entropy and to decode the meaning of the various data obtained, to a part of syntactic element, particularly mate with arbitrary shape relevant syntactic element of encoding, carry out the decoded partial decoding of h of entropy (as conversion decoding, predict and compensate the inverse operation namely asking residual error computing, DPCM decodes, single order and higher difference decoding, map decoding, runs decoding, index is decoded), the matched position of resolving and obtain after partial decoding of h and the arbitrary shape match parameter such as matched size or its variant, non-matched pixel value or its variant are exported to follow-up arbitrary shape matching and decoding step, resolve the every other syntactic element that obtains as mate residual error, the match pattern of remaining matching way and match parameter with arbitrary shape, the entropy decoded result of the relevant syntactic element such as coding result of various common technology exports data (i.e. entropy decode result) exports to remaining matching way follow-up and various common technology is decoded and restructing operation step, particularly, according to resolving the information that obtains or add the result of the characteristic of current C U and contiguous CU being carried out to analysis and evaluation according to resolving the information obtained from compressed bit stream data from compressed bit stream data, determine whether arbitrary shape matching and decoding is carried out to current C U, the example of the way of described analysis and evaluation: the way of foundation or the some decoded results with reference to contiguous CU, uses arbitrary shape matching and decoding mode and other decoding processes first to do one to current C U and takes turns part pre decoding and the way of evaluation part pre decoding result, if determine to carry out arbitrary shape matching and decoding to current C U, then order performs next step, otherwise, skip to step 3),

2) arbitrary shape matching and decoding step; Arbitrary shape matching and decoding mode and history pixel (reconstructed pixel) working area (also referred to as reference pixel sample value collection) that completed decoding is used to carry out the computing of arbitrary shape matching and decoding to current C U; The input of described arbitrary shape matching and decoding computing is resolved from compressed bit stream data in step 1) and I(I >=0 obtained of decoding) to (matched position, matched size) or its variant and J(J >=0) the arbitrary shape match parameter such as individual non-matched pixel value or its variant; Described matched position is used to represent from where copying in reference pixel sample value collection mates reference sample value and the position being pasted the coupling current sample of current C U; Obviously, coupling current sample is the duplicate of coupling with reference to sample value, and both are numerically equal; arbitrary shape matching and decoding mode (is called the current string of coupling or coupling current block with the pixel sample value string of variable-size or block, its position both also can represent with a linear address with 2 dimension coordinates) carry out arbitrary shape matching and decoding for unit, described coupling forms coupling string (block) in described reference pixel sample value collection with reference to sample value, be called coupling reference strings (block), its position is also both also can represent with a linear address with 2 dimension coordinates, thus in any form fit decoding process, described matched position both also can use the linear address that mates reference strings (block) and the difference of linear address of mating current string (block) by 2 dimension coordinates of coupling reference strings (block) and the difference of 2 dimension coordinates mating current string (block), be commonly referred to as displacement vector to represent, because the size (equaling the size of mating current string (block)) of coupling reference strings (block) is variable, so also need another to be called, the variable parameter of matched size is to represent its size, in the occasion of coupling string, matched size is exactly generally the length of coupling string, in the occasion of match block, the size of a block generally needs to represent with multiple parametric component, such as, block is the situation of rectangle, and matched size generally contains width and the height (sometimes also referred to as long and wide) of two components and rectangle, again such as, block is the situation of parallelogram, and matched size generally also contains the end and the height of two components and parallelogram, again such as, block is trapezoidal situation, and matched size generally contains three components and trapezoidal upper base, goes to the bottom and height, described non-matched pixel value is the direct pixel value that parsing and decoding obtain from compressed bit stream data and is pasted the position of pixel value in the current decoding of current C U, and described non-matched pixel value is not present in described reference pixel sample value collection usually, if to resolve from compressed bit stream data and decoding obtains is not the position of non-matched pixel value itself but non-matched pixel value, so just the position of non-matched pixel value is outputted to subsequent step 3) to calculate imitative matched pixel value, the output of described arbitrary shape matching and decoding computing is non-matched pixel value (or its position) described in described coupling current sample (numerically equal with reference to sample value with coupling) adds, the complete arbitrary shape matching and decoding that described coupling current sample and the described non-matched pixel value (or its position) that may exist all just form described current C U altogether exports, I(I >=0 finally obtained) individual coupling string (block) and J(J >=0) individual non-matched pixel value is seamless spliced just in time forms the current complete square shape of CU of decoding, when carrying out the computing of arbitrary shape matching and decoding, first determine to mate with reference to sample value position residing in history pixel (reconstructed pixel) temporary storage module (also referred to as reference pixel sample value collection) and the shape size determining the matching area (matching area can be coupling string or match block) be made up of with reference to sample value coupling, then the coupling reference sample value in the described matching area of described shape size is copied from described position, and described coupling is moved and pastes the position of mating current sample in current decoding with reference to sample value, restore coupling current sample, the output of this module is the described coupling current sample of all recoveries and possible non-matched pixel value,

3) remaining matching way and various common technology are decoded and restructing operation step, complete all the other decoding and restructing operations of current C U, other matching ways and various common technology are carried out to the various parameter and variable of mating CU in string (or match block) or current decoding in current decoding, as infra-frame prediction, inter prediction, Block-matching, palette mates, bar mates, microlith mates, microlith String matching, imitative coupling sample value prediction interpolation, inverse transformation, inverse quantization, corresponding to prediction residual and the compensation (namely getting the inverse operation of residual error computing) of mating residual error, predict and compensate (inverse operation namely asking residual error computing), DPCM, single order and higher difference, map, the distance of swimming, index, block-eliminating effect filtering, sample value adaptive equalization (Sample Adaptive Offset), decoding and restructing operation, other matching ways described refer in coding/decoding method and are different from described step 2) matching way of arbitrary shape matching and decoding that carries out, the input of this step is above-mentioned steps 1) the entropy decoding as mated residual error of the every other syntactic element that exports exports data and step 2) output and as described in coupling current sample add may exist as described in non-matched pixel value (or its position), and from the reference pixel in described reference pixel sample value collection, the output of this step is reconstructed pixel (comprising Perfect Reconstruction pixel and partial reconfiguration pixel in various degree), described reconstructed pixel is put among described reference pixel sample value collection, as the computing of follow-up arbitrary shape matching and decoding, remaining matching way and the decoding of various common technology and the reference pixel required for restructing operation step, described Perfect Reconstruction pixel is also the last output of this coding/decoding method.

The general synoptic diagram of code device of the present invention as shown in Figure 9.Whole code device forms by with lower module:

1) pixel pre-treatment and arbitrary shape coupling coding module: pixel pre-treatment and arbitrary shape coupling coding are implemented to inputted video image pixel value, produce and the matched position of output matching string (block) and matched size or its variant, non-matched pixel value or its variant, coupling residual sum other arbitrary shapes coupling coding result parameter and variable, the example of described pixel pre-treatment: sample value quantizes, color quantizing and based on the pixel cluster of color, to come accurately with palette and index or the color of approximate representation input original pixels, described matched position is used to represent and mates coupling reference pixel value that current pixel value matches the where variable parameter in history pixel (reconstructed pixel) temporary storage module (also referred to as reference pixel sample value collection) in described current C U, described arbitrary shape coupling coded system (is called the current string of coupling or coupling current block with the pixel sample value string of variable-size or block, its position both also can represent with a linear address with 2 dimension coordinates) encode for unit carries out coupling, the coupling that described coupling reference pixel value is formed in described reference pixel sample value collection is gone here and there or match block, be called coupling reference strings or coupling reference block, its position is also both also can represent with a linear address with 2 dimension coordinates, thus in described arbitrary shape coupling coded system, described matched position both also can use the linear address that mates reference strings (block) and the difference of linear address of mating current string (block) by 2 dimension coordinates of coupling reference strings (block) and the difference of 2 dimension coordinates mating current string (block), be commonly referred to as displacement vector to represent, because the size (equaling the size of mating current string (block)) of coupling reference strings (block) is variable, so also need another to be called, the variable parameter of matched size is to represent its size, in the occasion of coupling string, matched size is exactly generally the length of coupling string, in the occasion of match block, the size of a block generally needs to represent with multiple parametric component, such as, block is the situation of rectangle, and matched size generally contains width and the height (sometimes also referred to as long and wide) of two components and rectangle, again such as, block is the situation of parallelogram, and matched size generally also contains the end and the height of two components and parallelogram, again such as, block is trapezoidal situation, and matched size generally contains three components and trapezoidal upper base, goes to the bottom and height, described non-matched pixel value is in described reference pixel sample value collection, do not find input original pixels sample value or its pixel sample value through pre-treatment of coupling according to predetermined matching criterior, described non-matched pixel value is an input original pixels sample value or its pixel sample value through pre-treatment, so also can represent with its position in current C U, described non-matched pixel value also can be similar to by the imitative coupling sample value damaging or can't harm calculated with one, if matched size is 1, this coupling string (block) also can process as non-matched pixel value, described coupling residual error is input original pixels sample value and described difference of mating reference pixel value, if the predetermined matching criterior of arbitrary shape coupling coded system is definitely accurate harmless coupling, then described coupling residual error is zero, namely this arbitrary shape coupling coded system does not just mate residual error as output, if the predetermined matching criterior of a kind of arbitrary shape coupling coded system approximate damages coupling, then described coupling residual error may be non-vanishing, the situation that another kind damages coupling first carries out sample value quantification to input original pixels sample value, color quantizing or the pixel cluster based on color, the pre-treatment of the color of approximate representation input original pixels is carried out with palette and index, and then carry out arbitrary shape coupling coding, in this situation, because sample value quantizes, color quantizing or the pixel cluster based on color, the pixel that palette and index represent damages, even if arbitrary shape coupling coding itself is harmless, described coupling residual error (namely inputting original pixels sample value and described difference of mating reference pixel value) also may be non-vanishing, the example of described other arbitrary shapes coupling coding result: match pattern, uses coupling string to form the rule of arbitrary shape block, the type of the variant of match parameter, the computation rule of matched position, the computation rule of matched size, etc., the result of described current C U being carried out to the most fundamental sum essence of arbitrary shape coupling coding is I(I >=0) individual coupling goes here and there (block) and J(J >=0) individual non-matched pixel value, export I to (matched position, matched size) and J non-matched pixel value, described I coupling string (block) and J non-matched pixel are worth the complete square shape of seamless spliced just in time formation present encoding CU,

2) remaining matching way and various common technology are encoded and reconstructed module: to input video pixel, various parameter and variable implement other matching ways and various common technology, as infra-frame prediction, inter prediction, Block-matching, palette mates, bar mates, microlith mates, microlith String matching, imitative coupling sample value prediction interpolation, conversion, quantize, inverse transformation, inverse quantization, corresponding to prediction residual and the compensation (namely getting the inverse operation of residual error computing) of mating residual error, predict and ask residual error, DPCM, single order and higher difference, map, the distance of swimming, index, block-eliminating effect filtering, sample value adaptive equalization (Sample Adaptive Offset), coding and restructing operation and entropy code computing, remaining matching way described refers in code device and is different from described module 1) matching way of the arbitrary shape implemented coupling coding, the input of this module is described module 1) output, original described input video pixel and the reference pixel from described reference pixel sample value collection, the output of this module is reconstructed pixel value (comprising Perfect Reconstruction pixel and partial reconfiguration pixel in various degree) and the compressed bit stream containing any form fit coding result and other coding results, described reconstructed pixel value puts into reconstructed reference pixel value temporary storage module, as the reference pixel value required for follow-up arbitrary shape coupling encoding operation, remaining matching way and various common technology coding and restructing operation, described compressed bit stream is the last output of this code device, contains corresponding coding/decoding method and carries out decode and whole syntactic elements, particularly matched position required for reconstructing and the syntactic element such as matched size or its variant, non-matched pixel value or its variant,

3) history pixel (reconstructed pixel) temporary storage module: also referred to as reference pixel sample value collection, the temporary pixel value reconstructed, namely the reconstructed pixel value in an encoding process to the starting position of matching area in CU in present encoding or present encoding (can be coupling string or match block), as the reference pixel value of current and follow-up arbitrary shape coupling coding, reference pixel value required when being also used as to implement coding and the restructing operation of remaining matching way and various common technology to current and follow-up CU; Reconstructed pixel value just to reach after normally living through multiple working procedure and reconstructs completely; Therefore, the reconstructed pixel value stored in reference pixel sample value collection generally comprises the reconstructed pixel that namely different types of Perfect Reconstruction pixel and partial reconfiguration pixel in various degree only live through process portion; Different match patterns can adopt the reconstructed pixel of one species as with reference to pixel, and different types of reconstructed pixel also can be adopted as reference pixel; In same match pattern, different types of reconstructed pixel also can be adopted as reference pixel at different matched positions.

The general synoptic diagram of decoding device of the present invention as shown in Figure 10.Whole decoding device forms by with lower module:

1) bit stream data is resolved and partial decoding of h module: mate the matched position and matched size or its variant of encoding to what input containing arbitrary shape, the packed data of non-matched pixel value or its variant and the compressed bit stream of every other syntactic element packed data implement entropy decoding, and parse entropy and to decode the meaning of the various data obtained, to a part of syntactic element, particularly mate with arbitrary shape relevant syntactic element of encoding, carry out the decoded partial decoding of h of entropy (as conversion decoding, predict and compensate the inverse operation namely asking residual error computing, DPCM decodes, single order and higher difference decoding, map decoding, runs decoding, index is decoded), the matched position of resolving and obtain after partial decoding of h and the arbitrary shape match parameter such as matched size or its variant, non-matched pixel value or its variant are sent to arbitrary shape matching and decoding module, resolving other arbitrary shapes coupling coding results such as the coupling residual error that obtains, the match pattern of remaining matching way and match parameter, the entropy decoded result of syntactic element of coding result etc. of various common technology exports data (i.e. the result of entropy decoding) and is sent to remaining matching way and various common technology is decoded and reconstructed module, other example of described partial decoding of h: according to resolving the information that obtains or add the result of the characteristic of current C U and contiguous CU being carried out to analysis and evaluation according to resolving the information obtained from compressed bit stream data from compressed bit stream data, determine whether arbitrary shape matching and decoding is carried out to current C U, the example of the way of described analysis and evaluation: the way of foundation or the some decoded results with reference to contiguous CU, uses arbitrary shape matching and decoding mode and other decoding processes first to do one to current C U and takes turns part pre decoding and the way of evaluation part pre decoding result,

2) arbitrary shape matching and decoding module: implement the computing of arbitrary shape matching and decoding according to described matched position and the arbitrary shape match parameter such as matched size or its variant, non-matched pixel value or its variant; The input of this module is described module 1) resolving from compressed bit stream data and I(I >=0 that obtains of decoding of exporting) to (matched position, matched size) or its variant and J(J >=0) the arbitrary shape match parameter such as individual non-matched pixel value or its variant; Described matched position is used to represent from where copying in reference pixel sample value collection mates reference sample value and the position being pasted the coupling current sample of current C U; Obviously, coupling current sample is the duplicate of coupling with reference to sample value, and both are numerically equal; arbitrary shape matching and decoding mode (is called the current string of coupling or coupling current block with the pixel sample value string of variable-size or block, its position both also can represent with a linear address with 2 dimension coordinates) carry out arbitrary shape matching and decoding for unit, described coupling forms coupling string (block) in described reference pixel sample value collection with reference to sample value, be called coupling reference strings (block), its position is also both also can represent with a linear address with 2 dimension coordinates, thus in any form fit decoding process, described matched position both also can use the linear address that mates reference strings (block) and the difference of linear address of mating current string (block) by 2 dimension coordinates of coupling reference strings (block) and the difference of 2 dimension coordinates mating current string (block), be commonly referred to as displacement vector to represent, because the size (equaling the size of mating current string (block)) of coupling reference strings (block) is variable, so also need another to be called, the variable parameter of matched size is to represent its size, in the occasion of coupling string, matched size is exactly generally the length of coupling string, in the occasion of match block, the size of a block generally needs to represent with multiple parametric component, such as, block is the situation of rectangle, and matched size generally contains width and the height (sometimes also referred to as long and wide) of two components and rectangle, again such as, block is the situation of parallelogram, and matched size generally also contains the end and the height of two components and parallelogram, again such as, block is trapezoidal situation, and matched size generally contains three components and trapezoidal upper base, goes to the bottom and height, described non-matched pixel value is the direct pixel value that parsing and decoding obtain from compressed bit stream data and is pasted the position of pixel value in the current decoding of current C U, and described non-matched pixel value is not present in described reference pixel sample value collection usually, if to resolve from compressed bit stream data and decoding obtains is not the position of non-matched pixel value itself but non-matched pixel value, so just the position of non-matched pixel value is outputted to subsequent step 3) to calculate imitative matched pixel value, the output of described arbitrary shape matching and decoding computing is non-matched pixel value (or its position) described in described coupling current sample (numerically equal with reference to sample value with coupling) adds, the complete arbitrary shape matching and decoding that described coupling current sample and the described non-matched pixel value (or its position) that may exist all just form described current C U altogether exports, I(I >=0 finally obtained) individual coupling string (block) and J(J >=0) individual non-matched pixel value is seamless spliced just in time forms the current complete square shape of CU of decoding, when implementing arbitrary shape matching and decoding computing, first determine to mate with reference to sample value position residing in history pixel (reconstructed pixel) temporary storage module (also referred to as reference pixel sample value collection) and the shape size determining the matching area (matching area can be coupling string or match block) be made up of with reference to sample value coupling, then the coupling reference sample value in the described matching area of described shape size is copied from described position, and described coupling is moved and pastes the position of mating current sample in current decoding with reference to sample value, restore coupling current sample, the output of this module is the described coupling current sample of all recoveries and possible non-matched pixel value,

3) remaining matching way and various common technology are decoded and reconstructed module: implement other matching ways and various common technology to the various parameter and variable of mating CU in string (or match block) or current decoding in current decoding, as infra-frame prediction, inter prediction, inverse transformation, inverse quantization, corresponding to prediction residual and the compensation (namely getting the inverse operation of residual error computing) of mating residual error, DPCM, single order and higher difference, map, the distance of swimming, index, block-eliminating effect filtering, sample value adaptive equalization (Sample Adaptive Offset), decoding and restructing operation, other matching ways described refer in decoding device and are different from described module 2) matching way of arbitrary shape matching and decoding implemented, the output of this module is reconstructed pixel value (comprising Perfect Reconstruction pixel value and partial reconfiguration pixel value in various degree), described reconstructed pixel value puts into reconstructed reference pixel value temporary storage module, as the reference pixel value required for subsequent match decoding computing, remaining matching way and the decoding of various common technology and restructing operation, described Perfect Reconstruction pixel value is also the last output of this decoding device,

4) history pixel (reconstructed pixel) temporary storage module: also referred to as reference pixel sample value collection, the temporary pixel value reconstructed, namely the reconstructed pixel value in decode procedure to the starting position of matching area in CU in current decoding or current decoding (can be coupling string or match block), as the reference pixel value of current and follow-up arbitrary shape matching and decoding, reference pixel value required when being also used as to implement decoding and the restructing operation of remaining matching way and various common technology to current and follow-up CU; Reconstructed pixel value just to reach after normally living through multiple working procedure and reconstructs completely; Therefore, the reconstructed pixel value stored in reference pixel sample value collection generally comprises the reconstructed pixel that namely different types of Perfect Reconstruction pixel and partial reconfiguration pixel in various degree only live through process portion; Different match patterns can adopt the reconstructed pixel of one species as with reference to pixel, and different types of reconstructed pixel also can be adopted as reference pixel; In same match pattern, different types of reconstructed pixel also can be adopted as reference pixel at different matched positions.

Above provided diagram only illustrates basic conception of the present invention in a schematic way, then only the assembly relevant with the present invention is shown in graphic but not component count, shape and size when implementing according to reality is drawn, it is actual when implementing, and the kenel of each assembly, quantity and ratio can be a kind of change arbitrarily, and its assembly layout kenel also may be more complicated.

Below more implementation detail of the present invention and change.

an embodiment of coding method

The schematic flow sheet of an embodiment of coding method as shown in figure 11.The present embodiment comprises the steps:

3) to the LCU in present encoding, carry out the pixel cluster based on color, namely some very close color is merged into a color; For follow-up prediction steps and the needs mating step, pixel cluster can be carried out respectively or do not carry out pixel cluster;

4) to the process of the CU of 1 MxM pixel or carry out the intraframe predictive coding based on neighborhood pixels sample value and the interframe based on square (block) (namely between current encoded frame and former encoded frame) predictive coding without the pixel sample value of pixel cluster, be referred to as predictive coding, produce (1) prediction residual and (2) predictive mode and motion vector; Need in predictive coding to attempt multiple intra prediction mode and multiple inter-frame forecast mode (being referred to as predictive mode); Each predictive mode all can produce corresponding prediction residual and experience subsequent step 9), 10), 11), 12) computing;

5) carry out arbitrary shape coupling coding to the process of the CU of this MxM pixel or without the pixel sample value of pixel cluster, produce matched position p and the matched size s of (1) coupling residual sum (2) coupling string (block); Namely according to one or more prespecified match patterns, search in the history pixel set completing coding, that finds multiple optimum matched with the pixel sample value in present encoding CU one by one mates string or match block; This multiple arbitrary shape coupling string (block) is seamless spliced just in time forms the complete square shape of present encoding CU; If coupling is accurate harmless coupling, then without coupling residual error (coupling residual error is zero); If coupling approximate damages coupling, then mate residual error non-vanishing, general needs carry out follow-up optional transition coding step and quantization encoding step to coupling residual error, if but coupling residual error is very little, also coupling residual error can be considered as zero and do not carry out follow-up optional transition coding step and quantization encoding step; If this step have employed multiple prespecified match pattern, then each match pattern all can produce corresponding coupling residual error and experience subsequent step 9), 10), 11), 12) computing;

6) if the matched size s of coupling string (block) is greater than certain threshold value S(S general satisfaction: 2≤S≤6, if and S=0, so just only that the short coupling coding of secondary is carried out to non-matched pixel sample value, if S=1, so just only that the coupling string (block) being 1 to non-matched pixel sample value and matched size carries out the short coupling coding of secondary), then skip to step 8), otherwise, this coupling string (block) is short coupling string (block), the short coupling coding of following secondary is carried out to it: whether search can find coupling string (being called second degree matches string) in a represent pixel buffer memory (wherein each represent pixel and pixel sample value have a sequence number), if second degree matches string can be found, just export coupling sequence number (sequence number of first pixel sample value of the second degree matches string namely found in represent pixel buffer memory) and the matching length of second degree matches string, otherwise, export matched position and the matched size of the coupling string found in history pixel that original step 5) produces,

7) represent pixel buffer memory is upgraded; If the matched size s=0 of this short coupling string, namely any coupling is not all found to cause in described CU pixel sample value in present encoding to become non-matched pixel sample value in step 5) and step 6), so current non-matched pixel sample value will be determined whether to put into represent pixel buffer memory according to certain predetermined rule, and determine whether will delete certain pixel sample value from represent pixel buffer memory; A rule the simplest is exactly first-in first-out (FIFO) rule, namely all non-matched pixel sample values is put into represent pixel buffer memory all successively, after piling, just successively the non-matched pixel sample value put into the earliest is deleted; A more more complex rule is: when mating the matched size s≤S of string, just this short coupling being temporarily kept at a represent pixel spare area, but not putting into represent pixel buffer memory at once; When same short coupling appearance more than 2 times, then this short coupling is put into represent pixel buffer memory;

8) output of coupling coding is selected: if the matched size s=0 of this coupling string (block), then export non-matched pixel sample value, otherwise, if step 6) finds second degree matches string in represent pixel buffer memory, then export its coupling sequence number and matching length, otherwise, export matched position and the matched size of the coupling string (block) found in history pixel that original step 5) produces;

9) to step 4) and 5) prediction residual that produces and coupling residual error carry out transition coding, produces conversion coefficient; Transition coding is optional respectively, if namely carry out transition coding to prediction residual not reach better data compression effects, just transition coding is not carried out to prediction residual, if carry out transition coding to coupling residual error not reach better data compression effects, just transition coding is not carried out to coupling residual error;

10) quantization encoding is carried out to the conversion coefficient (having carried out the situation of transition coding in previous step) of prediction residual or prediction residual itself (not carrying out the situation of transition coding in previous step), produce and quantize predictive transformation coefficient or quantitative prediction residual error; Quantization encoding is carried out to the conversion coefficient (having carried out the situation of transition coding in previous step) or coupling residual error itself (not carrying out the situation of transition coding in previous step) that mate residual error, produces and quantize matched transform coefficient or quantize to mate residual error; In the situation of lossless coding, quantization encoding computing is identity of operation, does not produce the quantification effect of any reality;

11) to step 4), 9), 10) prediction-conversion-quantization encoding mode is (namely based on the coded system of prediction, referred to as prediction mode) result carry out the inverse operation of described prediction mode, to step 5), 9), 10) coupling-conversion-quantization encoding mode is (namely based on the coded system of coupling, referred to as matching way) result carry out the inverse operation of described matching way, obtain the multiple preliminary reconstruction pixel corresponding to multiple predictive mode and multiple match pattern of described CU, for subsequent step 13) rate-distortion cost calculation, and at subsequent step 16) determine the optimal dividing to present encoding LCU and the optimum code mode to each CU thereof after, all preliminary reconstruction pixels of optimal dividing are carried out to the reprocessing of block-eliminating effect filtering and pixel compensation computing, then the pixel of each reconstruct CU through reprocessing is put into history pixel (reconstructed pixel) working area, encode as subsequent match, the reference pixel of predictive coding and reconstruct, because coding may damage, reconstructed pixel might not equal to input original pixels,

12) to coupling coding outputs the i.e. quantization encoding result of match information and step 10) and quantitative prediction conversion coefficient, the quantitative prediction residual error of two predictive modes of the coding result of step 4) and motion vector, step 8), quantize matched transform coefficient and quantification and mate residual error and carry out entropy code and produce the bit rate of compressed bit stream; Entropy code also comprises predictive mode, motion vector, match pattern, matched position, matched size, coupling sequence number, matching length, non-matched pixel sample value, quantitative prediction conversion coefficient, quantitative prediction residual error, quantizes matched transform coefficient, quantizes differential coding or index coding that the sample value of coupling residual error constant entropy coded object first carries out tieing up based on 1 dimension or 2 contiguous sample value, to remove the correlation between sample value, improve the efficiency of entropy code;

13) from bit rate or the bit rate estimated value of original pixels, multiple preliminary reconstruction pixel and compressed bit stream thereof, calculate rate-distortion cost, select the optimum code mode of this CU (based on the coded system of mating still based on the coded system of prediction), Optimum Matching pattern or optimal prediction modes and the optimal compression bit stream data of this CU temporary according to rate-distortion performance;

14) judge whether to complete the coding to MxM pixel CU all in current division, if so, then order performs next step, otherwise skips to the coding that step 4) starts next MxM pixel CU;

15) judge whether that M has reached the minimum value preset, if so, then order performs next step, otherwise, M is reduced by half (degree of depth namely divided adds 1), then skips to the coding that step 4) starts all MxM pixel CU in the division of the next degree of depth;

16) from the coding result of the division of completed multiple different depth (CU of different size MxM pixel), select the division of the CU of an optimum MxM pixel according to rate-distortion performance and export the corresponding compressed bit stream data of this division;

17) judge whether the coding completing all LCU, if so, then terminate coding, otherwise get back to step 1), start the coding to next LCU.

an embodiment of coding/decoding method

The schematic flow sheet of an embodiment of coding/decoding method as shown in figure 12.The present embodiment comprises the steps:

1) the Partial shrinkage bit stream data of image is read in;

2) carry out the entropy decoding of a coding unit and CU, obtain header and the data message of this CU; What header adopted when comprising this CU coding is existing Predicting Technique or matching technique of the present invention; Which kind of arbitrary shape match pattern this CU have employed, or this CU infra-frame prediction that have employed based on neighborhood pixels sample value still have employed the inter prediction based on square; Whether carry out inverse transformation decoding; Entropy decoding also comprises predictive mode, motion vector, match pattern, matched position, matched size, coupling sequence number, matching length, non-matched pixel sample value, quantization transform coefficient, quantitative prediction residual error, the differential decoding based on 1 dimension or the contiguous sample values of 2 dimensions quantizing coupling residual error constant entropy decoder object or index decoding; In any form fit pattern, CU can by one or more coupling go here and there (block) splice and form, each coupling go here and there (block) subsequent step 6 will be experienced), 7), 8) computing;

3) to quantization transform coefficient or quantitative prediction residual error or quantize coupling residual error and carry out inverse quantization decoding, conversion coefficient or prediction residual or coupling residual error is produced;

4) if previous step produce be conversion coefficient, then carry out inverse transformation decoding, produce prediction residual or coupling residual error, otherwise, order perform next step;

5) if step 3) or 4) what produce is prediction residual, then carry out, based on the infra-frame prediction decoding of neighborhood pixels sample value or the interframe prediction decoding based on square, being referred to as prediction decoding, produce preliminary reconstructed pixel, then skip to step 9), otherwise order performs next step;

6) if step 2) entropy decoding obtains certain coupling sequence number and length, then and carry out short matching and decoding, from represent pixel stores, copy the pixel sample value obtaining short coupling string;

7) represent pixel buffer memory is upgraded; If matched size s=0, namely step 2) entropy decoding obtain a non-matched pixel sample value, so current non-matched pixel sample value will be determined whether to put into represent pixel buffer memory according to certain predetermined rule, and determine whether will delete certain pixel sample value from represent pixel buffer memory; A rule the simplest is exactly first-in first-out (FIFO) rule, namely non-matched pixel sample value is put into represent pixel buffer memory all successively, after piling, just successively the non-matched pixel sample value put into the earliest is deleted; A more more complex rule is: when mating the matched size s≤S of string, just this short coupling being temporarily kept at a represent pixel spare area, but not putting into represent pixel buffer memory at once; When same short coupling appearance more than 2 times, then this short coupling is put into represent pixel buffer memory;

8) arbitrary shape matching and decoding; According to step 2) the arbitrary shape match pattern that obtains, the coupling string (block) can with arbitrary shape is moved and pastes the current by the position of coupling string (block) of current decoding CU, add from step 3) or 4) coupling residual error, produce preliminary reconstructed pixel; The coupling string (block) of described arbitrary shape can be from step 2) non-matched pixel sample value, or go here and there from the short coupling of step 6), or from the coupling string (block) determined by these two match parameter of matched position p and matched size s in history pixel working area;

9) the preliminary reconstruction pixel of prediction mode or the preliminary reconstruction pixel of matching way are carried out to the reprocessing of block-eliminating effect filtering and pixel compensation computing, then whole reconstructed pixel of the current C U through reprocessing are put into history pixel (reconstructed pixel) working area, as the reference pixel of follow-up arbitrary shape matching and decoding and prediction decoding;

10) pixel of reconstruct CU is exported;

11) judge whether to complete the decoding to the compressed bit stream data of all CU in a LCU, if so, then order performs next step, otherwise gets back to step 1), start the decoding to next CU;

12) judge whether the decoding of the compressed bit stream data completed all LCU, if so, then terminate decoding, otherwise get back to step 1), start the decoding to next LCU.

an embodiment of code device

The module composition schematic diagram of an embodiment of code device as shown in figure 13.The present embodiment forms by with lower module:

1) memory module of the pixel of at least one coding unit and CU in the division of a kind of degree of depth in maximum coding unit in present encoding and LCU can be stored;

2) predictive coding module: input pixel is implemented based on the intraframe predictive coding of neighborhood pixels sample value and the inter prediction encoding based on square (block), and exports (1) prediction residual and (2) predictive mode and motion vector; This module can implement the coding under multiple intra prediction mode and multiple inter-frame forecast mode (being referred to as predictive mode);

3) based on the pixel cluster module of color: some very close color in input pixel is merged into a color, to improve the efficiency of follow-up coupling coding, find size larger, the coupling string (block) that error is less; In the situation that original color is fewer, or carry out the situation of ultrahigh quality coding and lossless coding, this module can be bypassed;

4) arbitrary shape match search coding module: to through or input pixel sample value without pixel cluster implement arbitrary shape coupling coding, and export (1) coupling residual error, (2) coupling string matched position p of (block) and matched size s, (3) coupling statistical information; This module can implement the coding under one or more match patterns;

5) short match search module: if the matched size s that certain coupling that arbitrary shape match search coding module searches is gone here and there is less than or equal to certain threshold value S, then implement the short match search of secondary and coding again, namely in a represent pixel cache module (wherein each pixel sample value has a sequence number), whether search can find coupling string (being called second degree matches string); If second degree matches string can be found, just export coupling sequence number (i.e. the sequence number of first pixel sample value in represent pixel cache module of second degree matches string) and the matching length of second degree matches string, otherwise, export matched position and the matched size of the coupling string found in history pixel that original arbitrary shape match search coding module exports; Coupling sequence number can represent with the bit number more less than matched position, thus can improve code efficiency; If S=0, be so just only that the short coupling coding of secondary is carried out to non-matched pixel sample value; If S=1, be so just only that the coupling string (block) being 1 to non-matched pixel sample value and matched size carries out the short coupling coding of secondary;

6) represent pixel cache module: store the represent pixel being used for short match search; This module can contain candidate's represent pixel buffer unit; This module can be empty; In this situation, just do not implement short match search, there is no described short match search module and represent pixel update module yet;

7) represent pixel update module: if matched size s=0, namely arbitrary shape match search coding module and short match search module all do not search any coupling and to cause in described CU pixel sample value in present encoding to become non-matched pixel sample value, non-matched pixel sample value so in present encoding is exactly the represent pixel of candidate, is placed into the candidate's represent pixel buffer unit in a represent pixel cache module; Decide current non-matched pixel sample value according to certain predetermined rule and whether become formal represent pixel immediately, or collect to a certain degree about the coupling statistical information data of this pixel sample value after become formal represent pixel again; When represent pixel memory space overflows, just need to determine how from represent pixel memory space, to delete certain pixel sample value; A method the simplest is exactly first-in first-out (FIFO), the first deletion of namely first putting into; A more more complex method decides to delete which pixel sample value according to the coupling statistical information data of each pixel sample value;

8) conversion module: transform operation is implemented, output transform coefficient to coupling residual sum prediction residual; To the screen picture pixel of some type, transform operation can not play the effect of packed data, and in this situation, just do not implement transform operation, namely conversion module is bypassed, direct output matching residual error or prediction residual;

9) quantization modules: to conversion coefficient (situation in the non-bypass of conversion module) or coupling residual error or prediction residual (situation conversion module is bypassed) implement quantization operations, output quantization predictive transformation coefficient or quantitative prediction residual error, and output quantization matched transform coefficient or quantification coupling residual error;

10) entropy code module: entropy code is implemented to result such as matched position, matched size, coupling sequence number, matching length, non-matched pixel sample value, quantization transform coefficient or the quantification coupling residual error etc. of coupling coding; To predictive coding result as predictive mode, motion vector, quantization transform coefficient or quantitative prediction residual error etc. implement entropy code, comprise and differential coding or the index coding of tieing up contiguous sample values based on 1 dimension or 2 are first implemented to the sample value of these entropy code objects;

11) reconstruct and post-processing module: the inverse operation implementing predictive coding module, conversion module, these three modules of quantization modules; Implement the inverse operation of arbitrary shape match search coding module, conversion module, these three modules of quantization modules, preliminary reconstruction goes out the pixel of described CU under multiple predictive mode and multiple match pattern, and preliminary reconstruction pixel is outputted to based on the optimal prediction modes of rate-distortion performance and match pattern and divide the calculating selecting module to be used for rate-distortion cost; Selecting after module determines the optimal dividing to present encoding LCU and the optimum code mode to each CU thereof and pattern thereof based on the optimal prediction modes of rate-distortion performance and match pattern and dividing, again the preliminary reconstruction pixel of each CU of optimal dividing is implemented to the reprocessing of block-eliminating effect filtering and pixel compensation computing, then the pixel of the reconstruct CU through reprocessing is put into history pixel (reconstructed pixel) temporary storage module, as follow-up CU predictive coding, coupling coding and reconstruct reference pixel;

12) history pixel (reconstructed pixel) temporary storage module: the reference pixel that this module provides predictive coding, coupling coding and reconstructs;

13) based on the optimal prediction modes of rate-distortion performance and match pattern with divides and selects module: in the division of the variant degree of depth of a coding LCU during each CU, the optimum code mode (based on the coded system of mating still based on the coded system predicted) of this CU, Optimum Matching pattern and optimal prediction modes is selected according to rate-distortion performance; Complete all different depths to a LCU division all CU coding after, select the degree of depth of the optimum of this LCU according to rate-distortion performance and corresponding to divide;

14) compressed bit stream output module: the coded compressed data of the optimal dividing selected is outputted to output video compressed bit stream.

an embodiment of decoding device

The module composition schematic diagram of an embodiment of decoding device as shown in figure 14.The present embodiment forms by with lower module:

1) Partial shrinkage bit stream data memory module;

2) entropy decoder module: implement entropy decoding to input compressed bit stream data, obtains header and the data message of current decoding CU; Entropy decoding also comprises predictive mode, motion vector, match pattern, matched position, matched size, coupling sequence number, matching length, non-matched pixel sample value, quantization transform coefficient, quantitative prediction residual error, the differential decoding based on 1 dimension or the contiguous sample values of 2 dimensions quantizing coupling residual error constant entropy decoder object or index decoding; Entropy decoding also comprises the degree of depth parsing the division that current decoding LCU adopts when encoding from input compressed bit stream data, current decoding CU adopts when encoding is existing Predicting Technique or matching technique of the present invention, have employed which kind of arbitrary shape match pattern, or the infra-frame prediction that have employed based on neighborhood pixels sample value still have employed the information such as whether to be bypassed based on the inter prediction of square, transform operation; Under match pattern, the data message of described current decoding CU can containing one or more information of mating string (block);

3) inverse quantization module: implement inverse quantization computing, output transform coefficient or prediction residual or coupling residual error;

4) inverse transform module; If transform operation is not bypassed during coding, then implement inverse transformation decoding, prediction of output residual error or coupling residual error, otherwise, do not implement inverse transformation decoding, direct prediction of output residual error or coupling residual error;

5) prediction decoding module: prediction residual is implemented based on the infra-frame prediction decoding of neighborhood pixels sample value or the interframe prediction decoding based on square, obtains and export preliminary reconstruction pixel;

6) represent pixel cache module: the input of this module is coupling sequence number and the length of the short coupling that entropy decoder module exports; The coupling sequence number of short coupling is the sequence number of first pixel sample value in represent pixel cache module in short coupling; According to coupling sequence number and the length of input, this module exports the short coupling be made up of some represent pixel sample values and goes here and there;

7) represent pixel update module: the effect of this module upgrades the represent pixel sample value in represent pixel cache module;

8) arbitrary shape matching and decoding module: when the input of this module is from the matched position of coupling string (block) of entropy decoder module and size time, the function of this module is that the coupling string (block) obtaining being of a size of matched size is copied in the place of being specified by matched position from history pixel temporary storage module, then this coupling string (block) is moved and paste the current by the position of coupling string (block) of current decoding CU; When the input of this module is the short coupling string from represent pixel cache module time, the function of this module directly described short coupling string is moved and paste the current by the position of coupling string of current decoding CU; When the input of this module is the non-matched pixel sample value from entropy decoder module time, the function of this module directly described non-matched pixel sample value is moved and paste the current by the position of coupling string of current decoding CU; Another function of this module be paste put in place by the pixel of coupling string (block), add the coupling residual error from inverse transform module, obtain and export preliminary reconstructed pixel; Coupling residual error can be also all zero, does not at this moment just need the coupling residual error add operation described in implementing;

9) post-processing module: reprocessing preliminary reconstruction pixel being implemented to block-eliminating effect filtering and pixel compensation computing, then the reconstructed pixel through reprocessing is put into history pixel (reconstructed pixel) temporary storage module, as the reference pixel of follow-up arbitrary shape matching and decoding and prediction decoding;

10) history pixel temporary storage module: this module provides the reference pixel of arbitrary shape matching and decoding and prediction decoding;

11) reconstructed pixel output module: reconstructed pixel is the last output of decoding device.

the embodiment that there is non-matched pixel sample value is represented in compressed bit stream

In any form fit Code And Decode, do not find the situation of coupling, namely have the situation of non-matched pixel sample value, namely can represent with matched size s=0, can mate with oneself by matched position p=0(current pixel sample value oneself yet) represent.

some embodiments of matching criterior

Coupling can be harmless or accurate, also can damage or be similar to; If the candidate matches string of a pixel sample value in the set of history pixel is x=(s _n, s _n+1..., s _n+m-1), in present encoding CU, current location is y=(s by coupling string _c, s _c+1..., s _c+m-1), so this is m to the matching length of pixel sample value string, and matching performance can with a length-distortion cost (Length-Distortion Cost is called for short LDcost) function LDcost=f (m, | s _c-s _n|, | s _c+1-s _n+1| ..., | s _c+m-1-s _n+m-1|) represent; The simplest a kind of length-distortion cost function be LDcost=(MaxStringLength-m)+λ (| s _c-s _n|+| s _c+1-s _n+1|+... + | s _c+m-1-s _n+m-1|), wherein MaxStringLength is default maximum matching length, if 300, λ is Lagrange gene, as 0.25, for balancing the weight of matching length and coupling distortion; In more complicated length-distortion cost function, each pixel sample value error term | s _c+q-1-s _n+q-1| may have the multiplier factor of oneself, multiplier factor also can change with length m; The effect of length-distortion cost function is when searching for coupling string (block), is used for the matching performance of evaluate matches string (block), selects Optimum Matching string (block); If when searching for coupling string (block), mandatory provision all pixel sample values error term | s _c+q-1-s _n+q-1|=0, then what obtain is exactly harmless or exact matching, otherwise what obtain is exactly damage or approximate match.

some embodiments of the match pattern of arbitrary shape coupling

In match pattern, adopt pixel value vertical arrangement form, namely first for the pixel value in one 2 dimension region vertical arrangement, another row are rearranged after having arranged row, with string shape coupling form, namely mate current pixel value and mate the string that reference pixel value all forms pixel value, two strings have equal length, but not necessarily keep 2 identical dimension shapes;

Or

In match pattern, adopt the horizontal form of pixel value, namely the pixel value in one 2 dimension region first horizontally, another row is rearranged after having arranged a line, with string shape coupling form, namely mate current pixel value and mate the string that reference pixel value all forms pixel value, two strings have equal length, but not necessarily keep 2 identical dimension shapes;

Or

In match pattern, adopt pixel value vertical arrangement form, namely first for the pixel value in one 2 dimension region vertical arrangement, another row are rearranged after having arranged row, with 2 repair and maintenance shape coupling forms, namely mate current pixel value and mate reference pixel value and all form 2 dimension regions, two regions have equal size, and keep 2 identical dimension shapes; The row that described 2 dimension regions are formed by pixel value vertical arrangement formed, centre except first row and last row respectively arranges identical height, the i.e. height in this region, first row aligns with the bottom in this region, and last row align with the top in this region; The height in region can be equal with the height of current C U, also can not wait with the height of current C U and get the value that another one is worth arbitrarily or derives from the height of current C U, as 1/2nd or 1/4th of the height of current C U; First row and last row also can respectively show identical height with centre, and namely equal the height in this region, at this moment, this region is rectangle or the square of an arbitrary size; In this match pattern, matching area is 2 dimension regions, matched size generally needs to represent with two or more parametric component, the special case that can derive from other parameters cataloged procedure and decode procedure or variable at the height of matching area (as equal the height of current C U or its 1/2nd), matched size also can only represent with a parametric component;

Or

In match pattern, adopt the horizontal form of pixel value, namely the pixel value in one 2 dimension region first horizontally, another row is rearranged after having arranged a line, with 2 repair and maintenance shape coupling forms, namely mate current pixel value and mate reference pixel value and all form 2 dimension regions, two regions have equal size, and keep 2 identical dimension shapes; Described 2 dimension regions are made up of horizontal the formed row of pixel value, middle rows except the first row and last column has identical width, the i.e. width in this region, the first row is alignd with the right part in this region, and last column is alignd with the left part in this region; The width in region can be equal with the width of current C U, also can not wait with the width of current C U and get the value that another one is worth arbitrarily or derives from the width of current C U, as 1/2nd or 1/4th of the width of current C U; The first row and last column also can have identical width with middle rows, namely equal the width in this region, and at this moment, this region is rectangle or the square of an arbitrary size; In this match pattern, matching area is 2 dimension regions, matched size generally needs to represent with two or more parametric component, the special case that can derive from other parameters cataloged procedure and decode procedure or variable at the width of matching area (as equal the width of current C U or its 1/2nd), matched size also can only represent with a parametric component;

Or

In match pattern, adopt the horizontal form of pixel value, namely the pixel value in one 2 dimension region first horizontally, another row is rearranged after having arranged a line, with 2 repair and maintenance shape coupling forms, namely mate current pixel value and mate reference pixel value and all form 2 dimension regions, two regions have equal size, and keep 2 identical dimension shapes; Described 2 dimension regions are made up of horizontal the formed row of pixel value, and there is identical width all provisional capitals; The left end point of all row can form a vertical straight line, and at this moment, this region is a rectangle or square, and also can form a non-perpendicular oblique line, at this moment, this region is a parallelogram;

Or

In match pattern, adopt the horizontal form of pixel value, namely the pixel value in one 2 dimension region first horizontally, another row is rearranged after having arranged a line, with 2 repair and maintenance shape coupling forms, namely mate current pixel value and mate reference pixel value and all form 2 dimension regions, two regions have equal size, and keep 2 identical dimension shapes; Described 2 dimension regions are made up of horizontal the formed row of pixel value, and the left end point of all row forms straight line, and the right endpoint of all row forms another straight line; These two straight lines can be parallel, and at this moment, this region is a parallelogram, also can be nonparallel, at this moment, this region be one trapezoidal;

In one width present image, same match pattern can be all used by all CU, also different match patterns can be used adaptively by different CU.

the embodiment of the matched position of matching area

The matched position of matching area (region can be string or block) is first coupling current sample the and mate relative position that in reference zone, first mates between reference sample value in coupling current region.

the embodiment 1 of the representation of matched position

Image is 2 dimension data, and coupling string is the special case of match block, and a dimension value namely in 2 dimensions is the special case of 1; So matched position is the natural number variable of one 2 dimension in essence, can represent (X, Y) with a natural number; But mathematically, the maximum of X and Y is represented respectively with Xmax and Ymax, then any one 2 dimension natural number can change into natural number L:L=X+W*Y of one 1 dimension to (X, Y), and wherein W is a natural number being greater than Xmax preset; L also can have another kind of computational methods: L=Y+H*X, and wherein H is a natural number being greater than Ymax preset; Obviously, also can calculate (X, Y) from L: the remainder and the business that with L divided by W, X and Y are just division respectively; Equally, X and Y is the quotient and the remainder of L divided by H respectively; Therefore, matched position can represent (X, Y) with a natural number; Also can represent with a natural number L, natural number W or H that (X, Y) and L can be determined according to a kind of rule in cataloged procedure and decode procedure by changes mutually; A kind of account form of described matched position (X, Y) is the 2 dimension coordinate (Xs of coupling with reference to sample value ₁, Y ₁) with the 2 dimension coordinate (X mating current sample ₂, Y ₂) difference add (picture traverse, picture altitude), and X ₁and X ₂all be less than picture traverse, Y ₁and Y ₂all be less than picture altitude, therefore X and Y is less than 2 times of picture traverse and 2 times of picture altitude respectively; The example of described W as can be seen here: 2 times of picture traverse, the example of described H: 2 times of picture altitude.

the embodiment 2 of the representation of matched position

Present image is divided into CU(or LCU one by one), each CU(or LCU) be endowed a sequence number, sequence number is started from scratch and is increased progressively one by one; Each pixel value in present image is also endowed a numbering (or claiming linear address), and numbering is started from scratch and increased progressively one by one; Putting in order of pixel value numbering is first from CU(or LCU) arrangement sequence number is little, CU(or LCU of a sequence number) whole pixel values all arranged after, rearrange CU(or LCU of Next Sequence) pixel value; And at CU(or LCU) inner, pixel sample value is with vertical format (or horizontal format is line by line) arrangement by column; The matched position of matching area be first coupling reference pixel value in coupling reference zone linear address (namely numbering) with to mate in current region the difference that first mates the linear address (namely numbering) of current pixel value.

the embodiment of the representation of limited matched position

If coupling is defined in reference to the position of sample value and mates current sample and have same horizontal level or upright position, one so in matched position (X, Y), is had to be zero (or other constants); With a natural number and one, matched position just can represent that level or the marker bit of upright position represent.

the embodiment of the representation of matched size

Matched size only has a parametric component, represents the length of coupling string;

Or

Matched size has a direct parameter component, as represented the number of the pixel value of coupling string (block), with a parametric component of deriving from other parameters cataloged procedure or decode procedure or variable (height or width as current C U), as represented height or the width of coupling string (block);

Or

Matched size has two parametric component, represents width and the height in the region of coupling string occupied by (block), these two parametric component can represent (X, Y) with a natural number, also can represent with a natural number L, natural number W or H that (X, Y) and L can be determined according to a kind of rule in cataloged procedure and decode procedure by one changes mutually, the example of described W: the 1) width of current C U, 2) horizontal range of pixel value to the right margin of current C U in the most upper left corner of current string (block) is mated, 3) maximum value possible of the width of the current string of coupling (block) determined by coupling current string (block) position in present image (or 2 dimension coordinates), 4) maximum value possible of the width of the coupling reference strings (block) determined by coupling reference strings (block) position in present image (or 2 dimension coordinates), 5) more than 3) and 4) smaller value in both, 6) maximum value possible of the width of the current string (block) determined by the region having completed coupling in position in current C U of the pixel value in the most upper left corner of the current string of coupling (block) and current C U, 7) by the maximum value possible of the width of the coupling reference strings (block) determined the restriction of coupling reference strings (block), 8) more than 6) and 7) smaller value in both, the example of described H: the 1) height of current C U, 2) vertical range of pixel value to the lower boundary of current C U in the most upper left corner of current string (block) is mated, 3) maximum value possible of the height of the current string of coupling (block) determined by coupling current string (block) position in present image (or 2 dimension coordinates), 4) maximum value possible of the height of the coupling reference strings (block) determined by coupling reference strings (block) position in present image (or 2 dimension coordinates), 5) more than 3) and 4) smaller value in both, 6) maximum value possible of the height of the current string (block) determined by the region having completed coupling in position in current C U of the pixel value in the most upper left corner of the current string of coupling (block) and current C U, 7) by the maximum value possible of the height of the coupling reference strings (block) determined the restriction of coupling reference strings (block), 8) more than 6) and 7) smaller value in both,

Or

Matched size has three parametric component, represents the size in the more complicated region of coupling string occupied by (block), as trapezoidal upper base, goes to the bottom and height; These three parametric component can represent with a natural number tlv triple (X, Y, Z), also can represent with a natural number L; 2 natural number W and H that (X, Y, Z) and L can be determined according to a kind of rule in cataloged procedure and decode procedure by one change mutually; Conversion process generally can be divided into two steps, the first step be natural number to the mutual conversion between (X, Y) and natural number U, second step is that natural number is to the mutual conversion between (U, Z) and natural number L.

the embodiment that pixel value represents with palette and index

Pixel value index k(0≤k < K of element in a palette containing K particular pixel values and palette) come accurately or approx to represent.Palette can be static also can be dynamically updated in cataloged procedure and decode procedure; In arbitrary shape matching process of the present invention and device, all places relating to " pixel value " can replace with " index " of pixel value; Such as, coupling current sample, coupling is with reference to sample value, and the coupling current sample index such as reference pixel sample value collection, coupling is with reference to sample value index, and reference pixel sample value indexed set etc. replaces.

different match patterns adopts the embodiment of different types of reconstructed pixel

There is the reconstructed pixel of two kinds of types: kind I and kind II, have 3 kinds of match patterns: Mode A, Mode B, pattern C; Mode A and Mode B adopt the reconstructed pixel of kind I, and pattern C adopts the reconstructed pixel of kind II; Namely, when carrying out arbitrary shape coupling coding or decoding to current C U, which kind of reconstructed pixel the match pattern according to current C U adopts as reference pixel according to following logical relation determination current C U:

(if match pattern==Mode A), adopt the reconstructed pixel of kind I as reference pixel;

(if match pattern==Mode B), adopt the reconstructed pixel of kind I as reference pixel;

(if match pattern==pattern C), adopt the reconstructed pixel of kind II as reference pixel;

Above symbol "==" expression " is equal to ".

in same match pattern, adopt the embodiment of different types of reconstructed pixel at different matched positions

Have the reconstructed pixel of two kinds of types: kind I and kind II, reference pixel sample value collection is divided into 3 part: part A, part B and part C; Adopt the reconstructed pixel of kind I when mating reference pixel value at part A and part B, and adopt the reconstructed pixel of kind II when mating reference pixel value at part C; Namely, when carrying out arbitrary shape coupling coding or decoding to current C U, which kind of reconstructed pixel the matched position according to current matching string (block) adopts as reference pixel according to following logical relation determination current matching string (block):

(if the coupling reference pixel value of being specified by matched position is positioned at part A), adopt the reconstructed pixel of kind I as reference pixel;

(if the coupling reference pixel value of being specified by matched position is positioned at part B), adopt the reconstructed pixel of kind I as reference pixel;

(if the coupling reference pixel value of being specified by matched position is positioned at part C), adopt the reconstructed pixel of kind II as reference pixel.

in compressed bit stream, sequence parameter set is containing the species number of reference pixel sample value collection and the embodiment of size thereof

The sequence parameter set of described compressed bit stream contains the syntactic element being loaded with following message:

, the species number P of reference pixel sample value collection, the size of the 1st kind reference pixel sample value collection, the 2nd kind reference pixel sample value collection size ..., P kind reference pixel sample value collection size,

Or, only have the size of the reference pixel sample value collection of a kind to need to write the situation of sequence parameter set (not having the size of the reference pixel sample value collection of the reference pixel sample value collection of other kinds or other kinds not need to write sequence parameter set), the sequence parameter set of described compressed bit stream contains the syntactic element being loaded with following message:

, reference pixel sample value collection size,

containing the packed data of match parameter of arbitrary shape coupling and the embodiment 1 of the compressed bit stream of other coding results

CU part in the described compressed bit stream mating coding result and other coding results containing arbitrary shape is made up of the syntactic element being loaded with following message:

CU head Part I, match pattern, CU head Part II, (matched position, matched size) or non-matched pixel sample value 1, (matched position, matched size) or non-matched pixel sample value 2, (matched position, matched size) or non-matched pixel sample value 3 ..., more (matched position, matched size) or non-matched pixel sample value, coupling residual error or sky, other coding results;

Except described this syntactic element of CU head Part I, it not is unique that the placement of every other institute's syntax elements in code stream puts in order, and can adopt any one predetermined rational order; Any one syntactic element also can be splitted into a few part, and described a few part can be concentrated and be placed on same place in code stream, also can be placed on places different in code stream respectively; Any some syntactic elements also can be merged into a syntactic element; Except described this syntactic element of CU head Part I, other syntactic elements also can not be present in the compressed bit stream data of certain CU;

Matched position in compressed bit stream, matched size, the match parameter such as non-matched pixel sample value, can be these match parameter itself, also can be the variants of these match parameter after various common technology coding such as predictive coding, coupling coding, transition coding, quantization encoding, DPCM, single order and higher difference coding, mapping code, Run-Length Coding, index coding etc.;

Described matched position only can have a parametric component, also can have two parametric component, or Further Division becomes three parametric component (horizontal component in the sequence number of such as CU and CU and vertical component) even four parametric component; Two parametric component can represent with a numerical value, also can represent with two numerical value; Three parametric component can represent with a numerical value, also can represent with two or three numerical value; Four parametric component can represent with a numerical value, also can represent by two or three or four numerical value;

Described matched size only can have a parametric component, also can have two, three even four parametric component; Two parametric component can represent with a numerical value, also can represent with two numerical value; Three parametric component can represent with a numerical value, also can represent with two or three numerical value; Four parametric component can represent with a numerical value, also can represent by two or three or four numerical value;

Described (matched position, matched size) and described non-matched pixel sample value can represent with different forms, also can represent with unified form.

containing the packed data of match parameter of arbitrary shape coupling and the embodiment 2 of the compressed bit stream of other coding results

CU head Part I, match pattern flag or sky, match pattern identification code or match pattern part identifer code or sky, CU head Part II, (matched position, matched size) or non-matched pixel sample value 1, (matched position, matched size) or non-matched pixel sample value 2, (matched position, matched size) or non-matched pixel sample value 3 ..., more (matched position, matched size) or non-matched pixel sample value, coupling residual error or sky, other coding results;

Described match pattern flag can value 1 or 0;

If described match pattern flag is empty, then hiddenly show that described match pattern flag gets a fixed value preset.

If described match pattern flag value 1, so also have this syntactic element of match pattern identification code (but not having this syntactic element of match pattern part identifer code) in compressed bit stream, described match pattern identification code can be got following code value and have following semanteme:

Code value	Semantic
		0	Current C U adopts match pattern 1
1	Current C U adopts match pattern 2
		2	Current C U adopts match pattern 3
…	……………………………………
		A-1	Current C U adopts match pattern A

If described match pattern flag value 0, so also has this syntactic element of match pattern part identifer code (but not having this syntactic element of match pattern identification code) in compressed bit stream, described match pattern part identifer code can get B(B < A) individual code value, the code value of described match pattern part identifer code itself can not determine which kind of match pattern current C U adopts completely, but must combine with the result of the characteristic of current C U and contiguous CU being carried out to analysis and evaluation and could determine which kind of match pattern current C U adopts completely, the example of the characteristic of described current C U and contiguous CU has: the coded system of contiguous CU, prediction or match pattern, motion vector, matched position, several or whole A kind match pattern is used first to do to current C U the characteristic that takes turns the result of part pre decoding, the characteristic of the contiguous decoding and reconstituting pixel sample value of CU and the part pre decoding reconstructed pixel sample value of current C U is as belonged to continuous tone content or non-continuous tone content, described match pattern part identifer code also may be empty, in this situation, carries out the result of analysis and evaluation to determine which kind of match pattern current C U adopts completely by the characteristic of current C U and contiguous CU.

containing the packed data of match parameter of arbitrary shape coupling and the embodiment 3 of the compressed bit stream of other coding results

CU head Part I, match pattern, palette or part palette or sky, represent pixel buffer memory or part represent pixel buffer memory or sky, CU head Part II, (matched position, matched size) or (the coupling sequence number of the short coupling of secondary, the matching length of the short coupling of secondary or sky) or non-matched pixel sample value 1, (matched position, matched size) or (the coupling sequence number of the short coupling of secondary, the matching length of the short coupling of secondary or sky) or non-matched pixel sample value 2, (matched position, matched size) or (the coupling sequence number of the short coupling of secondary, the matching length of the short coupling of secondary or sky) or non-matched pixel sample value 3, more (matched position, matched size) or (the coupling sequence number of the short coupling of secondary, the matching length of the short coupling of secondary or sky) or non-matched pixel sample value, coupling residual error or sky, other coding results,

Described palette produces in the pretreatment process of the color carrying out accurately or approx to represent input original pixels with palette and index; Multiple CU may share a palette constantly updated, so containing whole palette in the compressed bit stream of not necessarily each CU, and can only containing the part palette needing to upgrade or not containing palette;

Described represent pixel buffer memory for the short coupling of secondary, particularly to the short coupling of secondary of non-matched pixel value or non-matched pixel value index; Multiple CU may share a represent pixel buffer memory constantly updated, so containing whole represent pixel buffer memory in the compressed bit stream of not necessarily each CU, and can only containing the part represent pixel buffer memory needing to upgrade or not containing represent pixel buffer memory;

Described (the coupling sequence number of the short coupling of secondary, the matching length of the short coupling of secondary or sky) is the coupling sequence number that finds in described represent pixel buffer memory of short coupling and matching length; Only to do not find in reference pixel sample value collection or reference pixel sample value indexed set coupling or matched size be 1 pixel value or pixel value index carry out the short coupling of secondary and to encode and the matching length of the short coupling of secondary can only be that a fixed value preset is (as 1, represent that secondary short coupling only has a pixel value) special case, the syntactic element of the matching length of the short coupling of described secondary is empty, hidden show the matching length of the short coupling of described secondary get described in the fixed value that presets.

Described represent pixel buffer memory can be different from described palette, also can be equal to described palette.

the expansion of " pixel "

Large pixel is that some neighbors are regarded as one " composite pixel ".Figure 15 is some examples of large pixel, is described as follows.

A CU has 16x16 pixel (single pixel).

If adjacent 2 the single pixels of level form 1 large pixel, so 1 macro block just has 8x16 large pixel.To so large pixel, if pixel format is YUV4:2:2, so a large pixel just has 4 pixel component, i.e. 2 luminance Y component, 1 colourity U component, 1 colourity V component.

If vertically adjacent 2 single pixels form 1 large pixel, so 1 macro block just has 16x8 large pixel.

If adjacent 4 the single pixels of horizontal vertical form 1 large pixel, so 1 macro block just has 8x8 large pixel.To so large pixel, if pixel format is YUV4:2:0, so a large pixel just has 6 pixel component, i.e. 4 luminance Y component, 1 colourity U component, 1 colourity V component.

All apparatus and method relating to pixel sample value coupling Code And Decode in the present invention, can be defined as " large pixel " by the most basic element of coding wherein or decoding (being referred to as pixel sample value or sample value or pixel value) and becoming the apparatus and method relating to large pixel matching coding; Particularly, apparatus and method of the present invention, can be applicable to the Code And Decode of YUV4:2:2 pixel format and YUV4:2:0 pixel format.

Accompanying drawing explanation

Fig. 1 is coding method schematic flow sheet in prior art

Fig. 2 is coding/decoding method schematic flow sheet in prior art

Fig. 3 is the module composition schematic diagram of code device in prior art

Fig. 4 is the module composition schematic diagram of decoding device in prior art

When Fig. 5 is arbitrary shape coupling coding, for present encoding CU, in history pixel, search for multistage arbitrary shape coupling string

When Fig. 6 is arbitrary shape matching and decoding, after multistage arbitrary shape coupling string being copied from history pixel, paste present encoding CU

Fig. 7 is coding method schematic flow sheet of the present invention

Fig. 8 is coding/decoding method schematic flow sheet of the present invention

Fig. 9 is the module composition schematic diagram of code device of the present invention

Figure 10 is the module composition schematic diagram of decoding device of the present invention

Figure 11 is the schematic flow sheet of an embodiment of coding method

Figure 12 is the schematic flow sheet of an embodiment of coding/decoding method

Figure 13 is the module composition schematic diagram of an embodiment of code device

Figure 14 is the module composition schematic diagram of an embodiment of decoding device

Figure 15 is some examples of large pixel.

Claims

1. a method for encoding images, is characterized in that comprising the following steps:

Step 1) pixel characteristic analysis and evaluation, pre-treatment and coding mode selection step; Analysis and evaluation is carried out to the pixel sample value characteristic of current C U and contiguous CU, necessary pre-treatment is carried out to input pixel sample value, whether prejudge arbitrary shape coupling coded system is applicable to for encoding to current C U, this step is optional, namely can skip this step, directly enter next step;

Step 2) arbitrary shape coupling coding step, arbitrary shape coupling coded system and encoded history pixel (reconstructed pixel) working area (also referred to as reference pixel sample value collection) is used to carry out arbitrary shape coupling encoding operation to current C U, the input of described arbitrary shape coupling coding is input original pixels or its pixel through pre-treatment, described arbitrary shape mates the matched position and matched size or its variant that the output of encoding is coupling string (block), non-matched pixel value or its variant, and other arbitrary shapes of coupling residual sum mate coding result parameter and variable, described matched position is used to represent and mates coupling reference pixel value that current pixel value matches the where variable parameter in described reference pixel sample value collection in described current C U, described arbitrary shape coupling coded system (is called the current string of coupling or coupling current block with the pixel sample value string of variable-size or block, its position both also can represent with a linear address with 2 dimension coordinates) encode for unit carries out coupling, the coupling that described coupling reference pixel value is formed in described reference pixel sample value collection is gone here and there or match block, be called coupling reference strings or coupling reference block, its position is also both also can represent with a linear address with 2 dimension coordinates, thus in described arbitrary shape coupling coded system, described matched position both also can use the linear address that mates reference strings (block) and the difference of linear address of mating current string (block) by 2 dimension coordinates of coupling reference strings (block) and the difference of 2 dimension coordinates mating current string (block), be commonly referred to as displacement vector to represent, because the size (equaling the size of mating current string (block)) of coupling reference strings (block) is variable, so also need another to be called, the variable parameter of matched size is to represent its size, described non-matched pixel value is in described reference pixel sample value collection, do not find input original pixels sample value or its pixel sample value through pre-treatment of coupling according to predetermined matching criterior, described non-matched pixel value is an input original pixels sample value or its pixel sample value through pre-treatment, so also can represent with its position in current C U, described non-matched pixel value also can be similar to by the imitative coupling sample value damaging or can't harm calculated with one, if matched size is 1, this coupling string (block) also can process as non-matched pixel value, described coupling residual error is input original pixels sample value and described difference of mating reference pixel value, if the predetermined matching criterior of arbitrary shape coupling coded system is definitely accurate harmless coupling, then described coupling residual error is zero, namely this arbitrary shape coupling coded system does not just mate residual error as output, if the predetermined matching criterior of a kind of arbitrary shape coupling coded system approximate damages coupling, then described coupling residual error may be non-vanishing, the situation that another kind damages coupling first carries out sample value quantification to input original pixels sample value, color quantizing or the pixel cluster based on color, the pre-treatment of the color of approximate representation input original pixels is carried out with palette and index, and then carry out arbitrary shape coupling coding, in this situation, because sample value quantizes, color quantizing or the pixel cluster based on color, the pixel that palette and index represent damages, even if arbitrary shape coupling coding itself is harmless, described coupling residual error (namely inputting original pixels sample value and described difference of mating reference pixel value) also may be non-vanishing, the result of described current C U being carried out to the most fundamental sum essence of arbitrary shape coupling coding is I(I >=0) individual coupling goes here and there (block) and J(J >=0) individual non-matched pixel value, export I to (matched position, matched size) and J non-matched pixel value, described I coupling string (block) and J non-matched pixel are worth the complete square shape of seamless spliced just in time formation present encoding CU,

Remaining matching way of step 3) and various common technology are encoded and restructing operation step, complete all the other coding and restructing operations of current C U, the coding of other matching ways and various common technology and restructing operation and entropy code computing are carried out to the input original pixels of inputted video image, various parameter and variable; Remaining matching way described refers in coding method and is different from described step 2) matching way of the arbitrary shape that carries out coupling coding; The input of this step is above-mentioned steps 1) and step 2) output and input original pixels and the reference pixel from described reference pixel sample value collection; The output of this step is reconstructed pixel (comprising Perfect Reconstruction pixel and partial reconfiguration pixel in various degree) and the compressed bit stream containing any form fit coding result and other coding results; Described reconstructed pixel is put among described reference pixel sample value collection, as follow-up arbitrary shape coupling encoding operation, remaining matching way and various common technology coding and the reference pixel required for restructing operation step; Described compressed bit stream is the last output of this coding method, contains corresponding coding/decoding method and carries out decode and whole syntactic elements, particularly matched position required for reconstructing and the syntactic element such as matched size or its variant, non-matched pixel value or its variant.

2. a picture decoding method, is characterized in that comprising the following steps:

Step 1) compressed bit stream Data Analysis and partial decoding of h step; Entropy decoding is carried out to the matched position containing arbitrary shape coupling coding of input and the packed data of matched size or its variant, non-matched pixel value or its variant and the compressed bit stream of every other syntactic element packed data, and parse entropy and to decode the meaning of the various data obtained, to a part of syntactic element, particularly mate with arbitrary shape relevant syntactic element of encoding, carry out the decoded partial decoding of h of entropy; The matched position of resolving and obtain after partial decoding of h and the arbitrary shape match parameter such as matched size or its variant, non-matched pixel value or its variant are exported to follow-up arbitrary shape matching and decoding step; Resolve the every other syntactic element that obtains as mate residual error, the match pattern of remaining matching way and match parameter with arbitrary shape, the entropy decoded result of the relevant syntactic element such as coding result of various common technology exports data (i.e. entropy decode result) exports to remaining matching way follow-up and various common technology is decoded and restructing operation step; Particularly, according to resolving the information that obtains or add the result of the characteristic of current C U and contiguous CU being carried out to analysis and evaluation according to resolving the information obtained from compressed bit stream data from compressed bit stream data, determine whether arbitrary shape matching and decoding is carried out to current C U; If determine to carry out arbitrary shape matching and decoding to current C U, then order performs next step, otherwise, skip to step 3);

Step 2) arbitrary shape matching and decoding step; Arbitrary shape matching and decoding mode and history pixel (reconstructed pixel) working area (also referred to as reference pixel sample value collection) that completed decoding is used to carry out the computing of arbitrary shape matching and decoding to current C U; The input of described arbitrary shape matching and decoding computing is resolved from compressed bit stream data in step 1) and I(I >=0 obtained of decoding) to (matched position, matched size) or its variant and J(J >=0) the arbitrary shape match parameter such as individual non-matched pixel value or its variant; Described matched position is used to represent from where copying in reference pixel sample value collection mates reference sample value and the position being pasted the coupling current sample of current C U; arbitrary shape matching and decoding mode (is called the current string of coupling or coupling current block with the pixel sample value string of variable-size or block, its position both also can represent with a linear address with 2 dimension coordinates) carry out arbitrary shape matching and decoding for unit, described coupling forms coupling string (block) in described reference pixel sample value collection with reference to sample value, be called coupling reference strings (block), its position is also both also can represent with a linear address with 2 dimension coordinates, thus in any form fit decoding process, described matched position both also can use the linear address that mates reference strings (block) and the difference of linear address of mating current string (block) by 2 dimension coordinates of coupling reference strings (block) and the difference of 2 dimension coordinates mating current string (block), be commonly referred to as displacement vector to represent, because the size (equaling the size of mating current string (block)) of coupling reference strings (block) is variable, so also need another to be called, the variable parameter of matched size is to represent its size, described non-matched pixel value is the direct pixel value that parsing and decoding obtain from compressed bit stream data and is pasted the position of pixel value in the current decoding of current C U, and described non-matched pixel value is not present in described reference pixel sample value collection usually, if to resolve from compressed bit stream data and decoding obtains is not the position of non-matched pixel value itself but non-matched pixel value, so just the position of non-matched pixel value is outputted to subsequent step 3) to calculate imitative matched pixel value, the output of described arbitrary shape matching and decoding computing is non-matched pixel value (or its position) described in described coupling current sample (numerically equal with reference to sample value with coupling) adds, the complete arbitrary shape matching and decoding that described coupling current sample and the described non-matched pixel value (or its position) that may exist all just form described current C U altogether exports, I(I >=0 finally obtained) individual coupling string (block) and J(J >=0) individual non-matched pixel value is seamless spliced just in time forms the current complete square shape of CU of decoding, when carrying out the computing of arbitrary shape matching and decoding, first determine to mate with reference to sample value position residing in history pixel (reconstructed pixel) temporary storage module (also referred to as reference pixel sample value collection) and the shape size determining the matching area (matching area can be coupling string or match block) be made up of with reference to sample value coupling, then the coupling reference sample value in the described matching area of described shape size is copied from described position, and described coupling is moved and pastes the position of mating current sample in current decoding with reference to sample value, restore coupling current sample, the output of this module is the described coupling current sample of all recoveries and possible non-matched pixel value,

Remaining matching way of step 3) and various common technology are decoded and restructing operation step, complete all the other decoding and restructing operations of current C U, the various parameter and variable of mating CU in string (or match block) or current decoding in current decoding are carried out to decoding and the restructing operation of other matching ways and various common technology; Other matching ways described refer in coding/decoding method and are different from described step 2) matching way of arbitrary shape matching and decoding that carries out; The input of this step is above-mentioned steps 1) the entropy decoding as mated residual error of the every other syntactic element that exports exports data and step 2) output and as described in coupling current sample add may exist as described in non-matched pixel value (or its position), and from the reference pixel in described reference pixel sample value collection; The output of this step is reconstructed pixel (comprising Perfect Reconstruction pixel and partial reconfiguration pixel in various degree); Described reconstructed pixel is put among described reference pixel sample value collection, as the computing of follow-up arbitrary shape matching and decoding, remaining matching way and the decoding of various common technology and the reference pixel required for restructing operation step; Described Perfect Reconstruction pixel is also the last output of this coding/decoding method.

3. a picture coding device, is characterized in that comprising with lower module:

Module 1) pixel pre-treatment and arbitrary shape coupling coding module: to inputted video image pixel value implement pixel pre-treatment and arbitrary shape coupling coding, produce and the matched position of output matching string (block) and matched size or its variant, non-matched pixel value or its variant, coupling residual sum other arbitrary shapes coupling coding result parameter and variable, described matched position is used to represent and mates coupling reference pixel value that current pixel value matches the where variable parameter in history pixel (reconstructed pixel) temporary storage module (also referred to as reference pixel sample value collection) in described current C U, described arbitrary shape coupling coded system (is called the current string of coupling or coupling current block with the pixel sample value string of variable-size or block, its position both also can represent with a linear address with 2 dimension coordinates) encode for unit carries out coupling, the coupling that described coupling reference pixel value is formed in described reference pixel sample value collection is gone here and there or match block, be called coupling reference strings or coupling reference block, its position is also both also can represent with a linear address with 2 dimension coordinates, thus in described arbitrary shape coupling coded system, described matched position both also can use the linear address that mates reference strings (block) and the difference of linear address of mating current string (block) by 2 dimension coordinates of coupling reference strings (block) and the difference of 2 dimension coordinates mating current string (block), be commonly referred to as displacement vector to represent, because the size (equaling the size of mating current string (block)) of coupling reference strings (block) is variable, so also need another to be called, the variable parameter of matched size is to represent its size, described non-matched pixel value is in described reference pixel sample value collection, do not find input original pixels sample value or its pixel sample value through pre-treatment of coupling according to predetermined matching criterior, described non-matched pixel value is an input original pixels sample value or its pixel sample value through pre-treatment, so also can represent with its position in current C U, described non-matched pixel value also can be similar to by the imitative coupling sample value damaging or can't harm calculated with one, if matched size is 1, this coupling string (block) also can process as non-matched pixel value, described coupling residual error is input original pixels sample value and described difference of mating reference pixel value, if the predetermined matching criterior of arbitrary shape coupling coded system is definitely accurate harmless coupling, then described coupling residual error is zero, namely this arbitrary shape coupling coded system does not just mate residual error as output, if the predetermined matching criterior of a kind of arbitrary shape coupling coded system approximate damages coupling, then described coupling residual error may be non-vanishing, the situation that another kind damages coupling first carries out sample value quantification to input original pixels sample value, color quantizing or the pixel cluster based on color, the pre-treatment of the color of approximate representation input original pixels is carried out with palette and index, and then carry out arbitrary shape coupling coding, in this situation, because sample value quantizes, color quantizing or the pixel cluster based on color, the pixel that palette and index represent damages, even if arbitrary shape coupling coding itself is harmless, described coupling residual error (namely inputting original pixels sample value and described difference of mating reference pixel value) also may be non-vanishing, the result of described current C U being carried out to the most fundamental sum essence of arbitrary shape coupling coding is I(I >=0) individual coupling goes here and there (block) and J(J >=0) individual non-matched pixel value, export I to (matched position, matched size) and J non-matched pixel value, described I coupling string (block) and J non-matched pixel are worth the complete square shape of seamless spliced just in time formation present encoding CU,

Module 2) remaining matching way and various common technology coding and reconstructed module: the coding of other matching ways and various common technology and restructing operation and entropy code computing are implemented to input video pixel, various parameter and variable; Remaining matching way described refers in code device and is different from described module 1) matching way of the arbitrary shape implemented coupling coding; The input of this module is described module 1) output, original described input video pixel and the reference pixel from described reference pixel sample value collection; The output of this module is reconstructed pixel value (comprising Perfect Reconstruction pixel and partial reconfiguration pixel in various degree) and the compressed bit stream containing any form fit coding result and other coding results; Described reconstructed pixel value puts into reconstructed reference pixel value temporary storage module, as the reference pixel value required for follow-up arbitrary shape coupling encoding operation, remaining matching way and various common technology coding and restructing operation; Described compressed bit stream is the last output of this code device, contains corresponding coding/decoding method and carries out decode and whole syntactic elements, particularly matched position required for reconstructing and the syntactic element such as matched size or its variant, non-matched pixel value or its variant;

Module 3) history pixel (reconstructed pixel) temporary storage module: also referred to as reference pixel sample value collection, the temporary pixel value reconstructed, namely the reconstructed pixel value in an encoding process to the starting position of matching area in CU in present encoding or present encoding (can be coupling string or match block), as the reference pixel value of current and follow-up arbitrary shape coupling coding, reference pixel value required when being also used as to implement coding and the restructing operation of remaining matching way and various common technology to current and follow-up CU; Reconstructed pixel value just to reach after normally living through multiple working procedure and reconstructs completely; Therefore, the reconstructed pixel value stored in reference pixel sample value collection generally comprises the reconstructed pixel that namely different types of Perfect Reconstruction pixel and partial reconfiguration pixel in various degree only live through process portion; Different match patterns can adopt the reconstructed pixel of one species as with reference to pixel, and different types of reconstructed pixel also can be adopted as reference pixel; In same match pattern, different types of reconstructed pixel also can be adopted as reference pixel at different matched positions.

4. a picture decoding apparatus, is characterized in that comprising with lower module:

Module 1) bit stream data resolves and partial decoding of h module: entropy implemented to the compressed bit stream containing the arbitrary shape packed data and every other syntactic element packed data that mate matched position and matched size or its variant, non-matched pixel value or its variant of encoding inputted and decodes, and parse entropy and to decode the meaning of the various data obtained, to a part of syntactic element, particularly mate with arbitrary shape relevant syntactic element of encoding, carry out the decoded partial decoding of h of entropy; The matched position of resolving and obtain after partial decoding of h and the arbitrary shape match parameter such as matched size or its variant, non-matched pixel value or its variant are sent to arbitrary shape matching and decoding module; Resolving other arbitrary shapes coupling coding results such as the coupling residual error that obtains, the match pattern of remaining matching way and match parameter, the entropy decoded result of syntactic element of coding result etc. of various common technology exports data (i.e. the result of entropy decoding) and is sent to remaining matching way and various common technology is decoded and reconstructed module;

Module 2) arbitrary shape matching and decoding module: implement the computing of arbitrary shape matching and decoding according to described matched position and the arbitrary shape match parameter such as matched size or its variant, non-matched pixel value or its variant; The input of this module is described module 1) resolving from compressed bit stream data and I(I >=0 that obtains of decoding of exporting) to (matched position, matched size) or its variant and J(J >=0) the arbitrary shape match parameter such as individual non-matched pixel value or its variant; Described matched position is used to represent from where copying in reference pixel sample value collection mates reference sample value and the position being pasted the coupling current sample of current C U; arbitrary shape matching and decoding mode (is called the current string of coupling or coupling current block with the pixel sample value string of variable-size or block, its position both also can represent with a linear address with 2 dimension coordinates) carry out arbitrary shape matching and decoding for unit, described coupling forms coupling string (block) in described reference pixel sample value collection with reference to sample value, be called coupling reference strings (block), its position is also both also can represent with a linear address with 2 dimension coordinates, thus in any form fit decoding process, described matched position both also can use the linear address that mates reference strings (block) and the difference of linear address of mating current string (block) by 2 dimension coordinates of coupling reference strings (block) and the difference of 2 dimension coordinates mating current string (block), be commonly referred to as displacement vector to represent, because the size (equaling the size of mating current string (block)) of coupling reference strings (block) is variable, so also need another to be called, the variable parameter of matched size is to represent its size, described non-matched pixel value is the direct pixel value that parsing and decoding obtain from compressed bit stream data and is pasted the position of pixel value in the current decoding of current C U, and described non-matched pixel value is not present in described reference pixel sample value collection usually, if to resolve from compressed bit stream data and decoding obtains is not the position of non-matched pixel value itself but non-matched pixel value, so just the position of non-matched pixel value is outputted to subsequent step 3) to calculate imitative matched pixel value, the output of described arbitrary shape matching and decoding computing is non-matched pixel value (or its position) described in described coupling current sample (numerically equal with reference to sample value with coupling) adds, the complete arbitrary shape matching and decoding that described coupling current sample and the described non-matched pixel value (or its position) that may exist all just form described current C U altogether exports, I(I >=0 finally obtained) individual coupling string (block) and J(J >=0) individual non-matched pixel value is seamless spliced just in time forms the current complete square shape of CU of decoding, when implementing arbitrary shape matching and decoding computing, first determine to mate with reference to sample value position residing in history pixel (reconstructed pixel) temporary storage module (also referred to as reference pixel sample value collection) and the shape size determining the matching area (matching area can be coupling string or match block) be made up of with reference to sample value coupling, then the coupling reference sample value in the described matching area of described shape size is copied from described position, and described coupling is moved and pastes the position of mating current sample in current decoding with reference to sample value, restore coupling current sample, the output of this module is the described coupling current sample of all recoveries and possible non-matched pixel value,

Module 3) remaining matching way and the decoding of various common technology and reconstructed module: decoding and the restructing operation of other matching ways and various common technology are implemented to the various parameter of CU and variable in coupling string (or match block) in current decoding or current decoding; Other matching ways described refer in decoding device and are different from described module 2) matching way of arbitrary shape matching and decoding implemented; The output of this module is reconstructed pixel value (comprising Perfect Reconstruction pixel value and partial reconfiguration pixel value in various degree); Described reconstructed pixel value puts into reconstructed reference pixel value temporary storage module, as the reference pixel value required for subsequent match decoding computing, remaining matching way and the decoding of various common technology and restructing operation; Described Perfect Reconstruction pixel value is also the last output of this decoding device;

Module 4) history pixel (reconstructed pixel) temporary storage module: also referred to as reference pixel sample value collection, the temporary pixel value reconstructed, namely the reconstructed pixel value in decode procedure to the starting position of matching area in CU in current decoding or current decoding (can be coupling string or match block), as the reference pixel value of current and follow-up arbitrary shape matching and decoding, reference pixel value required when being also used as to implement decoding and the restructing operation of remaining matching way and various common technology to current and follow-up CU; Reconstructed pixel value just to reach after normally living through multiple working procedure and reconstructs completely; Therefore, the reconstructed pixel value stored in reference pixel sample value collection generally comprises the reconstructed pixel that namely different types of Perfect Reconstruction pixel and partial reconfiguration pixel in various degree only live through process portion; Different match patterns can adopt the reconstructed pixel of one species as with reference to pixel, and different types of reconstructed pixel also can be adopted as reference pixel; In same match pattern, different types of reconstructed pixel also can be adopted as reference pixel at different matched positions.

5. coding method according to claim 1, is characterized in that described each step is decomposed and is refined into following steps:

It is the maximum coding unit of NxN pixel and the original pixels of LCU that step 1) reads in a size;

Step 2) initial value of variable M is set to N; Described variable M represents that the size of a coding unit in the current division of described LCU and CU is MxM pixel;

Step 3) carries out the pixel cluster based on color to described LCU; For follow-up prediction steps and the needs mating step, pixel cluster can be carried out respectively or do not carry out pixel cluster;

Step 4) carries out the intraframe predictive coding based on neighborhood pixels sample value and the inter prediction encoding based on square to the process of the CU of 1 MxM pixel or without the pixel sample value of pixel cluster, be referred to as predictive coding, produce (1) prediction residual and (2) predictive mode and motion vector; Need in predictive coding to attempt multiple intra prediction mode and multiple inter-frame forecast mode (being referred to as predictive mode); Each predictive mode all can produce corresponding prediction residual and experience subsequent step 9), 10), 11), 12) computing;

Step 5) carries out arbitrary shape coupling coding to the process of described CU or without the pixel sample value of pixel cluster, produces matched position p and the matched size s of (1) coupling residual sum (2) coupling string (block); The process of described arbitrary shape coupling coding is: according to the match pattern that one or more are prespecified, search in the history pixel set completing coding, the optimum finding the pixel sample value in multiple and described CU to match one by one mate string (block); The square shape that the described CU of described multiple coupling string (block) seamless spliced formation is complete; If described coupling is accurate harmless coupling, then without described coupling residual error (described coupling residual error is zero); If described coupling approximate damages coupling, then described coupling residual error is non-vanishing, general needs carry out follow-up optional transition coding step and quantization encoding step to described coupling residual error, if but coupling residual error is very little, also described coupling residual error can be considered as zero and do not carry out follow-up optional transition coding step and quantization encoding step; If this step have employed multiple prespecified match pattern, then each match pattern all can produce corresponding coupling residual error and experience subsequent step 9), 10), 11), 12) computing;

If the matched size s of described coupling string (block) of step 6) one is greater than certain threshold value S, then skip to step 8), otherwise, coupling string (block) described in this is short coupling string (block), carries out the short coupling coding of following secondary: whether search can find coupling string (being called second degree matches string) in a represent pixel buffer memory (wherein each represent pixel and pixel sample value have a sequence number) to it; If second degree matches string can be found, just export coupling sequence number (i.e. the sequence number of first pixel sample value in described represent pixel buffer memory of described second degree matches string) and the matching length of described second degree matches string, otherwise, export matched position and the matched size of the described coupling string found in described history pixel that original step 5) produces; S=0 only carries out secondary short coupling coding to non-matched pixel sample value, S=1, is exactly that the coupling string (block) that is only 1 to non-matched pixel sample value and matched size carries out the short coupling of secondary and encodes;

Step 7) upgrades represent pixel buffer memory; If the matched size s=0 of short coupling string described in this, namely any coupling is not all found to cause in described CU pixel sample value in present encoding to become non-matched pixel sample value in step 5) and step 6), so described non-matched pixel sample value will be determined whether to put into described represent pixel buffer memory according to certain predetermined rule, and determine whether will delete certain pixel sample value from described represent pixel buffer memory;

Step 8) selects the output of coupling coding: if the matched size s=0 of coupling string (block) described in this, non-matched pixel sample value described in then exporting, otherwise, if step 6) finds described second degree matches string in described represent pixel buffer memory, then export coupling sequence number and the matching length of described second degree matches string, otherwise, export that original step 5) finds this described in the matched position of coupling string (block) and matched size;

Step 9) is to step 4) and 5) the described prediction residual that produces and described coupling residual error carry out transition coding, produces conversion coefficient; Transition coding is optional respectively, if namely carry out transition coding to described prediction residual not reach better data compression effects, just transition coding is not carried out to described prediction residual, if carry out transition coding to described coupling residual error not reach better data compression effects, just transition coding is not carried out to described coupling residual error;

Step 10) carries out quantization encoding to the described conversion coefficient (having carried out the situation of transition coding in previous step) of described prediction residual or described prediction residual itself (not carrying out the situation of transition coding in previous step), produces and quantizes predictive transformation coefficient or quantitative prediction residual error; Quantization encoding is carried out to the described conversion coefficient (having carried out the situation of transition coding in previous step) of described coupling residual error or described coupling residual error itself (not carrying out the situation of transition coding in previous step), produces and quantize matched transform coefficient or quantize coupling residual error; In the situation of lossless coding, quantization encoding computing is identity of operation, does not produce the quantification effect of any reality;

Step 11) is to step 4), 9), 10) prediction-conversion-quantization encoding mode is (namely based on the coded system of prediction, referred to as prediction mode) result carry out the inverse operation of described prediction mode, and to step 5), 9), 10) coupling-conversion-quantization encoding mode is (namely based on the coded system of coupling, referred to as matching way) result carry out the inverse operation of described matching way, obtain the multiple preliminary reconstruction pixel corresponding to multiple described predictive mode and multiple described match pattern of described CU, for subsequent step 13) rate-distortion cost calculation, and at subsequent step 16) determine the optimal dividing to described LCU and the optimum code mode to each CU thereof after, all preliminary reconstruction pixels of described optimal dividing are carried out to the reprocessing of block-eliminating effect filtering and pixel compensation computing, then the pixel of each reconstruct CU through reprocessing is put into history pixel (reconstructed pixel) working area, encode as subsequent match, the reference pixel of predictive coding and reconstruct,

Described quantitative prediction conversion coefficient, quantitative prediction residual error that the output of the described coupling coding that the described predictive mode that step 12) produces step 4) and motion vector, step 8) produce and step 10) produce, quantize matched transform coefficient and quantize coupling residual error carry out entropy code and produce the bit rate of compressed bit stream; Entropy code also comprises predictive mode, motion vector, match pattern, matched position, matched size, coupling sequence number, matching length, non-matched pixel sample value, quantitative prediction conversion coefficient, quantitative prediction residual error, quantizes matched transform coefficient, quantizes differential coding or index coding that the sample value of coupling residual error constant entropy coded object first carries out tieing up based on 1 dimension or 2 contiguous sample value, to remove the correlation between sample value, improve the efficiency of entropy code;

Step 13) is from the bit rate of the described original pixels of described CU, described multiple preliminary reconstruction pixel and compressed bit stream thereof or bit rate estimated value, calculate rate-distortion cost, select the optimum code mode of described CU (based on the coded system of mating still based on the coded system of prediction), Optimum Matching pattern or optimal prediction modes and the optimal compression bit stream data of temporary described CU according to rate-distortion performance;

Step 14) judges whether to complete the coding to MxM pixel CU all in current division, and if so, then order performs next step, otherwise skips to the coding that step 4) starts next MxM pixel CU;

Step 15) judges whether that M has reached the minimum value preset, if so, then order performs next step, otherwise, M is reduced by half (degree of depth namely divided adds 1), then skips to the coding that step 4) starts all MxM pixel CU in the division of the next degree of depth;

Step 16) from the coding result of the division of completed multiple different depth (CU of different size MxM pixel), select the division of the CU of an optimum MxM pixel according to rate-distortion performance and export the corresponding compressed bit stream data of this division;

Step 17) judge whether the coding completing all LCU, if so, then terminate coding, otherwise get back to step 1), start the coding to next LCU.

6. coding/decoding method according to claim 2, is characterized in that described each step is decomposed and is refined into following steps:

Step 1) reads in the Partial shrinkage bit stream data of image;

Step 2) carry out a coding unit and CU entropy decoding, obtain header and the data message of described CU; What header adopted when comprising described CU coding is prediction mode or matching way; Which kind of arbitrary shape match pattern described CU have employed, or the described CU infra-frame prediction that have employed based on neighborhood pixels sample value still have employed the inter prediction based on square; Whether carry out inverse transformation decoding; Entropy decoding also comprises predictive mode, motion vector, match pattern, matched position, matched size, coupling sequence number, matching length, non-matched pixel sample value, quantization transform coefficient, quantitative prediction residual error, the differential decoding based on 1 dimension or the contiguous sample values of 2 dimensions quantizing coupling residual error constant entropy decoder object or index decoding; Under match pattern, described CU can by one or more coupling go here and there (block) splice and form, each coupling go here and there (block) subsequent step 6 will be experienced), 7), 8) computing;

Step 3) is to described quantization transform coefficient or quantitative prediction residual error or quantize coupling residual error and carry out inverse quantization decoding, produces conversion coefficient or prediction residual or coupling residual error;

If what step 4) previous step produced is conversion coefficient, then carry out inverse transformation decoding, produce prediction residual or coupling residual error, otherwise order performs next step;

If step 5) step 3) or 4) what produce is prediction residual, then carry out based on the infra-frame prediction decoding of neighborhood pixels sample value or the interframe prediction decoding based on square, be referred to as prediction decoding, produce the preliminary reconstruction pixel of described prediction mode, then step 9) is skipped to, otherwise order performs next step;

If step 6) step 2) entropy described coupling string (block) of of obtaining of decoding represents by coupling sequence number and length, then carries out short matching and decoding, copy the pixel sample value obtaining short coupling string from a represent pixel stores;

Step 7) upgrades described represent pixel buffer memory; If the matched size s=0 of described coupling string (block), namely step 2) entropy decoding obtain a non-matched pixel sample value, so described non-matched pixel sample value will be determined whether to put into described represent pixel buffer memory according to certain predetermined rule, and determine whether will delete certain pixel sample value from described represent pixel buffer memory;

Step 8) arbitrary shape matching and decoding; According to step 2) the arbitrary shape match pattern that obtains, the coupling string (block) can with arbitrary shape is moved and pastes the current by the position of coupling string (block) of described CU, add from step 3) or 4) described coupling residual error, produce the preliminary reconstruction pixel of described matching way; The coupling string (block) of described arbitrary shape can be from step 2) described non-matched pixel sample value, or go here and there from the described short coupling of step 6), or from the coupling string (block) determined by these two match parameter of matched position p and matched size s in history pixel working area;

Step 9) carries out the reprocessing of block-eliminating effect filtering and pixel compensation computing to the preliminary reconstruction pixel of described prediction mode or the preliminary reconstruction pixel of described matching way, then the reconstructed pixel of the described CU through described reprocessing is put into history pixel (reconstructed pixel) working area, as the reference pixel of follow-up arbitrary shape matching and decoding and prediction decoding;

Step 10) exports the described reconstructed pixel of described CU;

Step 11) judges whether to complete the decoding to the compressed bit stream data of all CU in a LCU, and if so, then order performs next step, otherwise gets back to step 1), start the decoding to next CU;

Step 12) judges whether the decoding of the compressed bit stream data completed all LCU, if so, then terminates decoding, otherwise gets back to step 1), start the decoding to next LCU.

7. code device according to claim 3, is characterized in that described each decomposition module and is refined into lower module:

Module 1) memory module of the pixel of at least one coding unit and CU in the division of a kind of degree of depth in maximum coding unit in present encoding and LCU can be stored;

Module 2) predictive coding module: input pixel is implemented based on the intraframe predictive coding of neighborhood pixels sample value and the inter prediction encoding based on square (block), and exports (1) prediction residual and (2) predictive mode and motion vector; This module can implement the coding under multiple intra prediction mode and multiple inter-frame forecast mode (being referred to as predictive mode);

Module 3) based on the pixel cluster module of color: some very close color in described input pixel is merged into a color, to improve the efficiency of follow-up coupling coding, finds size larger, the coupling string (block) that error is less; In the situation that original color is fewer, or carry out the situation of ultrahigh quality coding and lossless coding, this module can be bypassed;

Module 4) arbitrary shape match search coding module: to through or described input pixel sample value without pixel cluster implement arbitrary shape coupling coding, and export (1) coupling residual error, the matched position p of (2) multiple coupling string (block) and matched size s, (3) coupling statistical information; This module can implement the coding under one or more match patterns;

Module 5) short match search module: if described arbitrary shape match search coding module search certain described in mate string matched size s be less than or equal to certain threshold value S, then implement the short match search of secondary and coding again, namely in a represent pixel cache module (wherein each pixel sample value has a sequence number), whether search can find coupling string (being called second degree matches string); If second degree matches string can be found, just export coupling sequence number (i.e. the sequence number of first pixel sample value in described represent pixel cache module of described second degree matches string) and the matching length of described second degree matches string, otherwise, export matched position p and the matched size s of the described coupling string found in history pixel that original described arbitrary shape match search coding module exports; Coupling sequence number can represent with the bit number more less than matched position, thus can improve code efficiency; S=0 only carries out secondary short coupling coding to non-matched pixel sample value, S=1, is exactly that the coupling string (block) that is only 1 to non-matched pixel sample value and matched size carries out the short coupling of secondary and encodes;

Module 6) represent pixel cache module: store the represent pixel being used for short match search; This module can contain candidate's represent pixel buffer unit; This module can be empty; In this situation, just do not implement short match search, there is no described short match search module and represent pixel update module yet;

Module 7) represent pixel update module: if the matched size s=0 of described coupling string, namely described arbitrary shape match search coding module and described short match search module all do not search any coupling and to cause in described CU pixel sample value in present encoding to become non-matched pixel sample value, so, non-matched pixel sample value is exactly the represent pixel of candidate, is placed into the candidate's represent pixel buffer unit in described represent pixel cache module; Described in deciding according to certain predetermined rule, whether non-matched pixel sample value becomes formal represent pixel immediately, or becomes formal represent pixel again after collecting the coupling statistical information data about described pixel sample value to a certain degree; When described represent pixel memory space overflows, just need to determine how from described represent pixel memory space, to delete certain pixel sample value;

Module 8) conversion module: transform operation is implemented to described coupling residual sum prediction residual, output transform coefficient; To the screen picture pixel of some type, transform operation can not play the effect of packed data, and in this situation, just do not implement transform operation, namely conversion module is bypassed, and directly exports described coupling residual error or described prediction residual;

Module 9) quantization modules: quantization operations is implemented to described conversion coefficient (situation in the non-bypass of conversion module) or described coupling residual error or prediction residual (situation conversion module is bypassed), output quantization predictive transformation coefficient or quantitative prediction residual error, and output quantization matched transform coefficient or quantification coupling residual error;

Module 10) entropy code module: to coupling sequence number and matching length, described non-matched pixel sample value, the described quantification matched transform coefficient of described multiple coupling string matched position of (block) and matched size, described second degree matches string or quantize coupling residual error etc. and implement entropy code; Entropy code is implemented to described predictive mode, motion vector, quantitative prediction conversion coefficient or quantitative prediction residual error etc., comprises and differential coding or the index coding of tieing up contiguous sample values based on 1 dimension or 2 are first implemented to the sample value of these entropy code objects;

Module 11) reconstruct and post-processing module: the inverse operation implementing described predictive coding module, conversion module, these three modules of quantization modules; Implement the inverse operation of described arbitrary shape match search coding module, conversion module, these three modules of quantization modules, preliminary reconstruction goes out the pixel of described CU under multiple predictive mode and multiple match pattern, and preliminary reconstruction pixel is outputted to based on the optimal prediction modes of rate-distortion performance and match pattern and divide the calculating selecting module to be used for rate-distortion cost; Selecting after module determines the optimal dividing to described LCU and the optimum code mode to each CU thereof and pattern thereof based on the optimal prediction modes of rate-distortion performance and match pattern and dividing, again the preliminary reconstruction pixel of each CU of described optimal dividing is implemented to the reprocessing of block-eliminating effect filtering and pixel compensation computing, then the pixel of the reconstruct CU through described reprocessing is put into history pixel (reconstructed pixel) temporary storage module, as follow-up CU predictive coding, coupling coding and reconstruct reference pixel;

Module 12) history pixel (reconstructed pixel) temporary storage module: the reference pixel that this module provides predictive coding, coupling coding and reconstructs;

Module 13) based on the optimal prediction modes of rate-distortion performance and match pattern with divides and selects module: in the division of the variant degree of depth of the described LCU that encodes during any one CU, select the optimum code mode (based on the coded system of mating still based on the coded system predicted) of described CU, Optimum Matching pattern and optimal prediction modes according to rate-distortion performance; Complete all different depths to described LCU division all CU coding after, select the degree of depth of the optimum of described LCU according to rate-distortion performance and corresponding to divide;

Module 14) compressed bit stream output module: the coded compressed data of the described optimal dividing selected is outputted to output video compressed bit stream.

8. decoding device according to claim 4, is characterized in that described each decomposition module and is refined into lower module:

Module 1) Partial shrinkage bit stream data memory module;

Module 2) entropy decoder module: entropy decoding is implemented to input compressed bit stream data, obtains header and the data message of current decoding CU; Entropy decoding also comprises the coupling sequence number to predictive mode, motion vector, match pattern, matched position, matched size, short coupling, the matching length of short coupling, non-matched pixel sample value, quantization transform coefficient, quantitative prediction residual error, the differential decoding based on 1 dimension or the contiguous sample values of 2 dimensions quantizing coupling residual error constant entropy decoder object or index decoding; Entropy decoding also comprises the degree of depth parsing the division that current decoding LCU adopts when encoding from described input compressed bit stream data, described current decoding CU adopts when encoding is prediction mode or matching way, have employed which kind of arbitrary shape match pattern, or the described current decoding CU infra-frame prediction that have employed based on neighborhood pixels sample value still have employed based on information the such as whether inter prediction of square, transform operation are bypassed; Under match pattern, the data message of described current decoding CU can containing one or more information of mating string (block);

Module 3) inverse quantization module: implement inverse quantization computing, output transform coefficient or prediction residual or coupling residual error;

Module 4) inverse transform module: if transform operation is not bypassed during coding, then implement inverse transformation decoding, prediction of output residual error or coupling residual error, otherwise, do not implement inverse transformation decoding, direct prediction of output residual error or coupling residual error;

Module 5) prediction decoding module: described prediction residual is implemented based on the infra-frame prediction decoding of neighborhood pixels sample value or the interframe prediction decoding based on square, obtains and export preliminary reconstruction pixel;

Module 6) represent pixel cache module: the input of this module is coupling sequence number and the length of the short coupling that described entropy decoder module exports; The coupling sequence number of short coupling is the sequence number of first pixel sample value in described represent pixel cache module in short coupling; According to described coupling sequence number and length, this module exports the short coupling be made up of some represent pixel sample values and goes here and there;

Module 7) represent pixel update module: the effect of this module upgrades the represent pixel sample value in described represent pixel cache module;

Module 8) arbitrary shape matching and decoding module: when the input of this module is from the matched position of described coupling string (block) of described entropy decoder module and size time, the function of this module is the pixel sample value that the coupling string (block) obtaining being of a size of described matched size is copied in the place of being specified by described matched position from a history pixel temporary storage module, then the pixel sample value of described coupling string (block) is moved and paste the current by the position of coupling string (block) of described current decoding CU; When the input of this module is the described short coupling string from described represent pixel cache module time, the function of this module directly the pixel sample value of described short coupling string is moved and paste the current by the position of coupling string of described current decoding CU; When the input of this module is the non-matched pixel sample value from entropy decoder module time, the function of this module directly described non-matched pixel sample value is moved and paste the current by the position of coupling string of described current decoding CU; Another function of this module is gone here and there in the pixel of (block) the described coupling put in place in stickup, adds the described coupling residual error from described inverse transform module, obtain and export preliminary reconstruction pixel; Described coupling residual error can be also all zero, does not at this moment just need the coupling residual error add operation described in implementing;

Module 9) post-processing module: reprocessing preliminary reconstruction pixel being implemented to block-eliminating effect filtering and pixel compensation computing, then the reconstructed pixel through described reprocessing is put into history pixel (reconstructed pixel) temporary storage module, as the reference pixel of follow-up arbitrary shape matching and decoding and prediction decoding;

Module 10) history pixel temporary storage module: this module provides the reference pixel of arbitrary shape matching and decoding and prediction decoding;

Module 11) reconstructed pixel output module: reconstructed pixel is the last output of decoding device.

9. coding method according to claim 1 or coding/decoding method according to claim 2 or code device according to claim 3 or decoding device according to claim 4 or coding method according to claim 5 or coding/decoding method according to claim 6 or code device according to claim 7 or decoding device according to claim 8, it is characterized in that: in any form fit Code And Decode, do not find the situation of coupling, namely there is the situation of non-matched pixel sample value, namely can represent with matched size s=0, also can mate with oneself by matched position p=0(current pixel sample value oneself) represent.

10. coding method according to claim 1 or coding/decoding method according to claim 2 or code device according to claim 3 or decoding device according to claim 4 or coding method according to claim 5 or coding/decoding method according to claim 6 or code device according to claim 7 or decoding device according to claim 8, it is characterized in that: matching criterior can be harmless or mate accurately, also can be damage or approximate coupling.

11. coding methods according to claim 1,2,3,4,5,6,7,8,9,10 or coding/decoding method or code device or decoding device, is characterized in that: the match pattern that described arbitrary shape coupling adopts has following characteristics:

Adopt pixel value vertical arrangement form, namely first for the pixel value in one 2 dimension region vertical arrangement, another row are rearranged after having arranged row, with string shape coupling form, namely mate current pixel value and mate the string that reference pixel value all forms pixel value, two strings have equal length, but not necessarily keep 2 identical dimension shapes;

Or

Adopt the horizontal form of pixel value, namely the pixel value in one 2 dimension region first horizontally, another row is rearranged after having arranged a line, with string shape coupling form, namely mate current pixel value and mate the string that reference pixel value all forms pixel value, two strings have equal length, but not necessarily keep 2 identical dimension shapes;

Or

Adopt pixel value vertical arrangement form, namely first for the pixel value in one 2 dimension region vertical arrangement, another row are rearranged after having arranged row, with 2 repair and maintenance shape coupling forms, namely mate current pixel value and mate reference pixel value and all form 2 dimension regions, two regions have equal size, and keep 2 identical dimension shapes; The row that described 2 dimension regions are formed by pixel value vertical arrangement formed, centre except first row and last row respectively arranges identical height, the i.e. height in this region, first row aligns with the bottom in this region, and last row align with the top in this region; The height in region can be equal with the height of current C U, also can not wait with the height of current C U and get the value that another one is worth arbitrarily or derives from the height of current C U, as 1/2nd or 1/4th of the height of current C U; First row and last row also can respectively show identical height with centre, and namely equal the height in this region, at this moment, this region is rectangle or the square of an arbitrary size; In this match pattern, matching area is 2 dimension regions, matched size generally needs to represent with two or more parametric component, in the special case that the height of matching area can be derived from other parameters cataloged procedure and decode procedure or variable, matched size also can only represent with a parametric component;

Or

Adopt the horizontal form of pixel value, namely the pixel value in one 2 dimension region first horizontally, another row is rearranged after having arranged a line, with 2 repair and maintenance shape coupling forms, namely mate current pixel value and mate reference pixel value and all form 2 dimension regions, two regions have equal size, and keep 2 identical dimension shapes; Described 2 dimension regions are made up of horizontal the formed row of pixel value, middle rows except the first row and last column has identical width, the i.e. width in this region, the first row is alignd with the right part in this region, and last column is alignd with the left part in this region; The width in region can be equal with the width of current C U, also can not wait with the width of current C U and get the value that another one is worth arbitrarily or derives from the width of current C U, as 1/2nd or 1/4th of the width of current C U; The first row and last column also can have identical width with middle rows, namely equal the width in this region, and at this moment, this region is rectangle or the square of an arbitrary size; In this match pattern, matching area is 2 dimension regions, matched size generally needs to represent with two or more parametric component, in the special case that the width of matching area can be derived from other parameters cataloged procedure and decode procedure or variable, matched size also can only represent with a parametric component;

Or

Adopt the horizontal form of pixel value, namely the pixel value in one 2 dimension region first horizontally, another row is rearranged after having arranged a line, with 2 repair and maintenance shape coupling forms, namely mate current pixel value and mate reference pixel value and all form 2 dimension regions, two regions have equal size, and keep 2 identical dimension shapes; Described 2 dimension regions are made up of horizontal the formed row of pixel value, and there is identical width all provisional capitals; The left end point of all row can form a vertical straight line, and at this moment, this region is a rectangle or square, and also can form a non-perpendicular oblique line, at this moment, this region is a parallelogram;

Or

Adopt the horizontal form of pixel value, namely the pixel value in one 2 dimension region first horizontally, another row is rearranged after having arranged a line, with 2 repair and maintenance shape coupling forms, namely mate current pixel value and mate reference pixel value and all form 2 dimension regions, two regions have equal size, and keep 2 identical dimension shapes; Described 2 dimension regions are made up of horizontal the formed row of pixel value, and the left end point of all row forms straight line, and the right endpoint of all row forms another straight line; These two straight lines can be parallel, and at this moment, this region is a parallelogram, also can be nonparallel, at this moment, this region be one trapezoidal;

12. coding methods according to claim 1,2,3,4,5,6,7,8,9,10,11 or coding/decoding method or code device or decoding device, is characterized in that: the matched position of the matching area (region can be string or block) of described arbitrary shape coupling is first coupling current sample the and mate relative position that in reference zone, first mates between reference sample value in coupling current region.

13. coding methods according to claim 1,2,3,4,5,6,7,8,9,10,11,12 or coding/decoding method or code device or decoding device, it is characterized in that: the matched position of described arbitrary shape coupling can represent (X, Y) with a natural number; Also can represent with a natural number L, natural number W or H that (X, Y) and L can be determined according to a kind of rule in cataloged procedure and decode procedure by changes mutually.

14. coding methods according to claim 1,2,3,4,5,6,7,8,9,10,11,12,13 or coding/decoding method or code device or decoding device, it is characterized in that: present image is divided into CU(or LCU one by one), each CU(or LCU) be endowed a sequence number, sequence number is started from scratch and is increased progressively one by one; Each pixel value in present image is also endowed a numbering (or claiming linear address), and numbering is started from scratch and increased progressively one by one; Putting in order of pixel value numbering is first from CU(or LCU) arrangement sequence number is little, CU(or LCU of a sequence number) whole pixel values all arranged after, rearrange CU(or LCU of Next Sequence) pixel value; And at CU(or LCU) inner, pixel sample value is with vertical format (or horizontal format is line by line) arrangement by column; The matched position of matching area be first coupling reference pixel value in coupling reference zone linear address (namely numbering) with to mate in current region the difference that first mates the linear address (namely numbering) of current pixel value.

15. coding methods according to claim 1,2,3,4,5,6,7,8,9,10,11,12,13,14 or coding/decoding method or code device or decoding device, it is characterized in that: coupling is defined in reference to the position of sample value and mates current sample and have same horizontal level or upright position, has one to be zero (or other constants) in the horizontal component of matched position and vertical component; With the variable of a parametric component and one, matched position can represent that level or the marker bit of upright position represent.

16. coding methods according to claim 1,2,3,4,5,6,7,8,9,10,11,12,13,14,15 or coding/decoding method or code device or decoding device, is characterized in that: the matched size of described arbitrary shape coupling has following representation:

Or

Matched size has a direct parameter component and a parametric component of deriving from other parameters cataloged procedure or decode procedure or variable (height or width as current C U);

Or

Matched size has two parametric component, represents width and the height in the region of coupling string occupied by (block); These two parametric component can represent (X, Y) with a natural number, also can represent with a natural number L; Natural number W or H that (X, Y) and L can be determined according to a kind of rule in cataloged procedure and decode procedure by one changes mutually;

Or

Matched size has three parametric component, represents the size in the more complicated region of coupling string occupied by (block); These three parametric component can represent with a natural number tlv triple (X, Y, Z), also can represent with a natural number L; 2 natural number W and H that (X, Y, Z) and L can be determined according to a kind of rule in cataloged procedure and decode procedure by one change mutually; Conversion process generally can be divided into two steps, the first step be natural number to the mutual conversion between (X, Y) and natural number U, second step is that natural number is to the mutual conversion between (U, Z) and natural number L.

17. coding methods according to claim 1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16 or coding/decoding method or code device or decoding device, is characterized in that: all or part pixel value of present image or current C U index k(0≤k < K of element in a palette containing K particular pixel values and palette) come accurately or approx to represent; The described index of pixel value is carried out described arbitrary shape coupling coding or decoding as object (replacing described pixel value); Described palette can be static also can be dynamically updated in cataloged procedure and decode procedure.

18. coding methods according to claim 1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17 or coding/decoding method or code device or decoding device, it is characterized in that: the reconstructed pixel having two kinds of types: kind I and kind II having 3 kinds of match patterns: Mode A, Mode B, pattern C; Mode A and Mode B adopt the reconstructed pixel of kind I, and pattern C adopts the reconstructed pixel of kind II; Namely, when carrying out arbitrary shape coupling coding or decoding to current C U, which kind of reconstructed pixel the match pattern according to current C U adopts as reference pixel according to following logical relation determination current C U:

Above symbol "==" expression " is equal to ".

19. coding methods according to claim 1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18 or coding/decoding method or code device or decoding device, it is characterized in that: the reconstructed pixel having two kinds of types: kind I and kind II, reference pixel sample value collection is divided into 2 part: part A and part B; Adopt the reconstructed pixel of kind I when mating reference pixel value at part A, and adopt the reconstructed pixel of kind II when mating reference pixel value at part B; Namely, when carrying out arbitrary shape coupling coding or decoding to current C U, which kind of reconstructed pixel the matched position according to current matching string (block) adopts as reference pixel according to following logical relation determination current matching string (block):

(if the coupling reference pixel value of being specified by matched position is positioned at part B), adopt the reconstructed pixel of kind II as reference pixel.

20. coding methods according to claim 1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19 or coding/decoding method or code device or decoding device, is characterized in that: the sequence parameter set of described compressed bit stream contains the syntactic element being loaded with following message:

, reference pixel sample value collection size ...

21. coding methods according to claim 1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20 or coding/decoding method or code device or decoding device, is characterized in that: be describedly made up of the syntactic element being loaded with following message containing the CU part in the compressed bit stream of arbitrary shape coupling coding result and other coding results:

22. coding methods according to claim 1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20,21 or coding/decoding method or code device or decoding device, is characterized in that: be describedly made up of the syntactic element being loaded with following message containing the CU part in the compressed bit stream of arbitrary shape coupling coding result and other coding results:

Described match pattern flag can value 1 or 0;

If described match pattern flag is empty, then hiddenly show that described match pattern flag gets a fixed value preset;

23. coding methods according to claim 1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20,21,22 or coding/decoding method or code device or decoding device, is characterized in that: be describedly made up of the syntactic element being loaded with following message containing the CU part in the compressed bit stream of arbitrary shape coupling coding result and other coding results:

Described (the coupling sequence number of the short coupling of secondary, the matching length of the short coupling of secondary or sky) is the coupling sequence number that finds in described represent pixel buffer memory of short coupling and matching length; Only to do not find in reference pixel sample value collection or reference pixel sample value indexed set coupling or matched size be 1 pixel value or pixel value index carry out the short coupling of secondary and to encode and the matching length of the short coupling of secondary can only be that a fixed value preset is (as 1, represent that secondary short coupling only has a pixel value) special case, the syntactic element of the matching length of the short coupling of described secondary is empty, hidden show the matching length of the short coupling of described secondary get described in the fixed value that presets;