CN103647969B - A kind of object-based Fast Fractal video compress and decompression method - Google Patents

Abstract

The present invention proposes a kind of object-based Fast Fractal video compress and decompression method, video object segmentation plane i.e. Alpha plane is obtained first with methods of video segmentation, start frame is used block dct transform coding, non-I frame is carried out block motion estimation/compensation coding, calculate to the pixel of the relevant sub-block in sub-block territory and father's block territory and with pixel quadratic sum, the pixel of fractional pixel interpolation value corresponding blocks and, pixel quadratic sum, carry out pre-search restrictive condition judgement, and utilize the asymmetric cross multi-level hexagonal point search algorithm of fraction pixel Block-matching and improvement to find most like match block in former frame search window, utilize Huffman compression coding iterated function system coefficient.Homographic solution compression process is: non-I frame carries out Huffman decoding and obtains iterated function system coefficient, carry out decoding based on macro block, calculate father's block territory be correlated with sub-block pixel and with pixel quadratic sum, the most successively each macro block in present frame is decoded, and square loop circuit filtering method is removed in utilization.

Description

A kind of object-based Fast Fractal video compress and decompression method

The application is invention entitled " a kind of object-based Fast Fractal video compress and decompression method " (application Number: 201110188771.4, applying date: on 07 06th, 2011) divisional application of application for a patent for invention.

Technical field

The invention belongs to the video compression coding field in signal processing, propose particular for a new generation's field of video encoding A kind of object-based Fast Fractal video compressing and encoding method, on the premise of ensureing picture quality, is greatly accelerated fractal The speed of Video coding and compression ratio.

Background technology

The concept encoded based on object (Object-Based, be called for short OB) is proposed by MPEG-4 standard the earliest, use based on The video compressing and encoding method of object makes the foreground object of each frame video and background object be able to independent carrying out and encode, can To improve compression ratio further, some new functions can be realized in decoding end, such as each object video simultaneously Independent transmission and the replacement of decoding, object and background, object-based video frequency searching, coding staff especially with respect to standard Method, can obtain better image quality on the border of object, because the border of object is mutually the most overlapping with the edge in image Closing, it is the part of coding difficulty.MPEG-4 proposes object-based video coding technical standard, in video analysis, with Object video is ultimate unit, and each Video Events and scene can be seen as by the static genus of semantic objects all in video Property (such as shape, color, texture) and dynamic attribute (exercise intensity, direction, rhythm) thereof combine.It is being basic with object The video analysis method of unit of analysis, meets mode of thinking and the visual characteristic of the mankind, eliminates unessential part in video Interference to video analysis (sees Liang Zhang.Object-based Method of important video clips Extraction[J].Journal of Beijing Information Science and Technology University, 2009,24 (4): 83-86), (see Bo Huang, Yujian Huang.A Scalable Object- Based Wavelet Coder [J] .Radio Communications Technology, 2009,35 (1): 35-38), (ginseng See Kassim Ashraf a, Zhao L F.Rate-scalable object-based wavelet codec with implicit shape coding[J].IEEE Transactions on Circuits and Systems for Video Technology, 2000,10 (7): 1,068 1079).Object-based video coding method can improve video compression coding Performance also makes it have more motility.

Fractal theory initially is proposed (to see the seventies in last century by MandelbrotB.Mandelbrot.The Fractal Geometry of Nature [M] .New York:W.H.Freeman and Company, 1982.).Fractal volume The Fundamentals of Mathematics of code are that iterated function system (IFS) is theoretical.First fractal image is used for interactive image compression by Barnsley (see Michael F.Barnsley, Alan D.Sloan.A better way to compress image [J] .Byte Magazine, 1988,13 (1): 215-233.).Jacqain proposes full automatic Fractal Image Compression Approach and (sees Arnaud E.Jacquin.A novel fractal blocking-coding technique for digital image [C].IEEE International Conference on Acoustics,Speech and Signal Processing, 1990,4:2225-2228.), (Arnaud E.Jacquin.Fractal image coding:a review [J] is seen .Proceeding of the IEEE, 1993,81 (10): 1451-1465.), the method uses mode based on image block Affine transformation with local replaces the affine transformation of the overall situation.Afterwards, Fisher utilizes quaternary tree to improve this method (to see Y.Fisher.Fractal Image Compression [J] .Fractals, 1994,2 (3): 347-361.), (see Y.Fisher,E.W.Jacobs.Image compression:A study the iterated transform method [J] .Signal Processing, 1992,29 (3), 251-263.), (see Y.Fisher.Fractal Image Compression:Theory and application to digital images[M].New York:Spring- Verlag, 1995,55-77.), substantially increase code efficiency, and become the main stream approach in current fractal image coding.

On this basis, some scholars and research worker are applied to the pressure of video sequence the method for Fractal Image Compression In contracting.Meiqing Wang etc. proposes and comprehensively (sees based on data cube and fractal image compression system based on frame Meiqing Wang,Choi-Hong Lai.A hybrid fractal video compression method[J] .Computers&Mathematics with Applications, 2005,50 (3-4): 611-621.), (see Meiqing Wang,Zhehuang Huang,Choi-Hong Lai.Matching search in fractal video compression and its parallel implementation in distributed computing Environments [J] .Applied Mathematical Modeling, 2006,30 (8): 677-687.), (see Meiqing Wang,Rong Liu,Choi-Hong Lai.Adaptive partition and hybrid method in fractal video compression[J].Computers&Mathematics with Applications,2006,51 (11): 1715-1726.).What the most classical and impact was bigger sees (C.S.Kim, R.C.Kim, S.U.Lee.Fractal coding of video sequence using circular prediction mapping and noncontractive interframe mapping[J].IEEE Transactions on Image Processing,1998,7(4):601- 605.).The method uses and is similar to the motion estimation/compensation technology that standard video coder method is used, and this process employs Time strong correlation between consecutive frame, achieves preferable effect to compression of video sequence.In CPM and NCIM, sub-block territory In each image block obtained by motion compensation by the father's block territory from consecutive frame formed objects.CPM and NCIM is between the two Maximum difference is that CPM needs possess convergence during decoding, and NCIM need not.But encode at circular prediction (CPM) in method, in order to ensure start frame through self iterative decoding can approximate convergence to original image, compression process Needing through complex transformations, search and iteration etc., compression time and picture quality are difficult to reach requirement.The most typical graftal The operand of picture and video-frequency compression method is very big, and coding rate is relatively slow, and the quality decoded has much room for improvement so that fractal pattern Also need to further improve with video-frequency compression method.

Applicant has applied for two patents about fractal image in April, 2010: a kind of based on fractal Video compression coding-decoding method (201010167243.6CN101860753A) and the compression of one object-based fractal video are compiled Coding/decoding method (201010167235.1CN101827268A).The present invention is different from document disclosed above to be: 1) make use of pre- Search qualifications；2) make use of fraction pixel Block-matching；3) make use of the multi-level hexagonal of asymmetric cross of improvement Point search algorithm；4), in decoding, make use of square loop filtering.Therefore, coding efficiency has had the biggest improving.

Summary of the invention

The present invention proposes a kind of object-based Fast Fractal video compress and decompression method, divides first with video Segmentation method obtains video object segmentation plane i.e. Alpha plane, start frame uses block dct transform coding, non-I frame is carried out block Motion estimation/compensation encodes, calculate to the pixel of the relevant sub-block in sub-block territory and father's block territory and with pixel quadratic sum, in fraction pixel The pixel of interpolation corresponding blocks and, pixel quadratic sum, carry out pre-search restrictive condition judgement, and utilize in former frame search window point Number block of pixels coupling and the asymmetric cross multi-level hexagonal point search algorithm improved find most like match block, profit With Huffman compression coding iterated function system coefficient.Homographic solution compression process is: non-I frame carries out Huffman decoding and obtains Iterated function system coefficient, carries out decoding based on macro block, calculate father's block territory be correlated with sub-block pixel and with pixel quadratic sum, so After successively each macro block in present frame is decoded, and utilize remove square loop circuit filtering method.

A kind of object-based Fast Fractal video-frequency compression method, comprises the following steps:

Step one: utilize automatic video object segmentation algorithm automatically to split video sequence, obtains regarding of each frame Frequently Object Segmentation plane i.e. Alpha plane, determines object video region to be encoded according to the video object segmentation plane obtained； The frame being compressed is called present frame, the most encoded former frame built of laying equal stress on of present frame is called reference frame；

Step 2: first determine whether whether start frame is I frame, if I frame, first carries out non-overlapping copies to this I frame and consolidates The block of sizing divides, and each image block is respectively adopted I frame intraimage method based on block dct transform, to this Two field picture carries out separately encoded and decoding, forwards step 10 to；Otherwise, step 3 is forwarded to；Described I frame be video sequence start frame or Person's video sequence only carries out the picture frame of intraframe coding；Block in described piece of dct transform uses fixed size pattern；

Step 3: if present frame is non-I frame, before carrying out Block-matching, first present frame is divided into fixed size The macro block of non-overlapping copies, then calculate these macro blocks and the pixel of fritter obtained through tree-shaped division and, pixel quadratic sum, with And the most encoded of present frame lay equal stress in the former frame built i.e. reference frame, according to setting all macro blocks that step-length divides and through tree Shape divide the pixel of fritter obtained and, pixel quadratic sum, calculate simultaneously fractional pixel interpolation value corresponding blocks pixel and, pixel Quadratic sum, to reduce the double counting during Block-matching；Forward step 4 to；The collection that described present frame is all pieces is collectively referred to as sub-block Territory；The collection of all pieces of described former frame is collectively referred to as father's block territory；

Step 4: to the image block being presently processing i.e. current block, utilizes Alpha plane to differentiate the region of this image block Attribute；If this block is not in the object video region of present encoding, this block is not processed；If this block is all within currently In the object video region of coding, proceed to step 5；If the partial pixel of this block is in the object video region of present encoding, Partial pixel in the object video region of present encoding, does not needs individual processing, proceeds to step 9；If all of macro block all Processed complete, then forward step 10 to；The described not block in the object video region of present encoding is referred to as external block, described entirely Portion's all blocks in the object video region of present encoding are referred to as internal block, and described partial pixel is not at the video pair of present encoding As the block in region is boundary block；

Step 5: successively all macro blocks of present frame are encoded, first grand to this in the search window in father's block territory Block carries out Block-matching；In carrying out the sub-block matching process with father's block, the position of sub-block is as the initiating searches point of father's block, father's block Size identical with the size of sub-block, forward step 6 to；

Step 6: utilize the asymmetric cross multi-level hexagonal point search algorithm and fraction pixel block improved Join, search out optimal matching error: utilize the asymmetry search algorithm search improved, then search fractional pixel interpolation value pair Answer the RMS point at fritter, find the RMS point of minimum, forward step 7 to；

Step 7: pre-search restrictive condition judges: for specific sub-block, limits if meeting pre-search with father's block respective value Condition, then forward step 8 to；The most directly preserve current iterated function system coefficient i.e. IFS coefficient, proceed to step 4 coding Next macro block；

Step 8: if matching error RMS is less than the threshold gamma starting to set, preserve current iterated function system coefficient I.e. IFS coefficient, proceeds to step 4 and encodes next macro block；Otherwise, according to tree, this block is divided successively, and to each Divide the fritter obtained, utilize the asymmetric cross multi-level hexagonal point search of fraction pixel Block-matching and improvement to calculate Method, calculates matching error RMS respectively, if RMS is less than setting threshold gamma, then stops dividing and recording this fritter IFS coefficient, turns Enter step 4 and encode next macro block；Otherwise continuing to divide, until current block to be divided into smallest blocks set in advance, recording IFS Coefficient；Proceed to step 4 and encode next macro block；Described search window is the rectangular search region in reference frame；Described IFS coefficient Including father's block position, (x, y) with scale factor s, displacement factor o；If all of macro block of present frame is the most encoded complete, then turn To step 10；

Step 9: individual processing boundary block, the boundary block of present frame only belonging in boundary block and internal block in father's block Search coupling, the pixel Criterion of Selecting of current block and father's block is: only calculate the object video being positioned at present encoding in current block Pixel value in region, only carries out Block-matching in the same video subject area of reference frame；For in father's block, if with currently A certain pixel in father's block of the opposite position of block falls within this object video region, then use original pixel value, otherwise, use it It belongs to the meansigma methods of this object video area pixel and replaces；Return step 5 to process；

Step 10: all IFS coefficients carry out Huffman coding, reduces the statistical redundancy of IFS coefficient data；Judge to work as Whether front frame is last frame, if last frame terminates coding；Otherwise, return step 2 and continue with next frame image.

Described one object-based Fast Fractal video-frequency compression method, the video sequence of process is yuv format, the most right Above-mentioned ten steps of each employing in 3 components process.

Described step 6 mid score block of pixels is mated, including three below step:

1) pixel in region of search in reference frame is carried out interpolation and forms one relative to the pixel in integer position more High-resolution region；

2) carry out integer pixel in interpolation region and optimal coupling is found in half-pixel position search；

3) current block is substituted by the affine transformation of match block.

The asymmetric cross multi-level hexagonal point search algorithm improved in described step 6, in H.264 Asymmetrical hexagonal-shaped algorithm, the improvement of this algorithm is mainly reflected in following three points:

1) starting point prediction:

It is not involved with multi-reference frame based on fractal video coding algorithm, and macro block and sub-block have different big Little, therefore utilize three kinds of modes to carry out starting point prediction:

A) spatial domain median prediction: take a left side for current sub-block, upper, the motion vector intermediate value of right adjacent block is that predicted motion is vowed Amount；

B) initial point prediction: according to temporal correlation, makes motion vector value for (0,0)；

C) neighboring reference frame prediction: utilize the MV of correspondence position block in previous reference frame to be predicted in proportion；

2) threshold value jump condition during asymmetric cross template search:

The sub-block of fractal image and error matching criterior R of father's block are formula (3), (4), (5).The choosing that varies in size according to block Selecting different threshold values, asymmetric cross template search is complete, selects optimal match point and carries out follow-up masterplate as new starting point Coupling；

3) end condition is shifted to an earlier date:

Feature according to fractal coding algorithm will terminate being divided into two kinds of situations in advance: one is at non-homogeneous multi-level hexagon In lattice point whole pixel motion search procedure, in addition to the end condition in advance of this algorithm itself, in order to reduce search complexity such as Really optimum point is positioned at hexagonal centre, can stop search；Two is to use tree-shaped partition structure based on fractal video coding algorithm.

In described step 7 pre-search restrictive condition be following form wherein, b_iFor the pixel value of sub-block, a_iPicture for father's block Element value, s is the scale factor in fractal image, and o is displacement factor, and | | a | | represents two-dimentional norm, i.e. | | a | |=(| a₁|²+|a₂ |²+···+|a_n|²)^1/2:

$\mathrm{RMS}=\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{(s\·a_i+o-b_i)}^2$

$=\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{(s\·a_i+\frac{1}{n}[\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{b}_i-s\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{a}_i]-b_i)}^2$

$=\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{((a_i-\frac{\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{a}_i}{n})\·\frac{[n\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{a}_i{b}_i-\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{a}_i\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{b}_i]}{[n\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{a}_i^2-{\left(\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{a}_i\right)}^2]}+\frac{\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{b}_i}{n}-b_i)}^2$

$=\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{((a_i-\stackrel{\‾}{a})\·\frac{[\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{a}_i{b}_i-n\stackrel{\‾}{a}\stackrel{\‾}{b}]}{[\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{a}_i^2-n{\stackrel{\‾}{a}}^2]}+\stackrel{\‾}{b}-b_i)}^2$

$=\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{((a_i-\stackrel{\‾}{a})\·\frac{\underset{i=1}{\overset{n}{\mathrm{\Σ}}}(b_i-\stackrel{\‾}{b})(a_i-\stackrel{\‾}{a})}{{\left|\right|a_i-\stackrel{\‾}{a}\left|\right|}^2}+\stackrel{\‾}{b}-b_i)}^2$

$={\left|\right|b_i-\stackrel{\‾}{b}\left|\right|}^2\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{(\frac{{(a}_i-\stackrel{\‾}{a})}{\left|\right|a_i-\stackrel{\‾}{a}\left|\right|}\·\frac{\underset{i=1}{\overset{n}{\mathrm{\Σ}}}(b_i-\stackrel{\‾}{b})(a_i-\stackrel{\‾}{a})}{\left|\right|b_i-\stackrel{\‾}{b}\left|\right|\left|\right|a_i-\stackrel{\‾}{a}\left|\right|}-\frac{b_i-\stackrel{\‾}{b}}{\left|\right|b_i-\stackrel{\‾}{b}\left|\right|})}^2---\left(1\right)$

Allow $\hat{a}=\frac{(a_i-\stackrel{\‾}{a})}{\left|\right|a_i-\stackrel{\‾}{a}\left|\right|},$ $\hat{b}=\frac{b_i-\stackrel{\‾}{b}}{\left|\right|b_i-\stackrel{\‾}{b}\left|\right|},$ And understand ${\left|\right|\hat{a}\left|\right|}^2=1,$ ${\left|\right|\hat{b}\left|\right|}^2=1,$ Then R can derive as follows:

$\mathrm{RMS}={\left|\right|b_i-\stackrel{\‾}{b}\left|\right|}^2\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{(\hat{a}\·\underset{i=1}{\overset{n}{\mathrm{\Σ}}}\hat{b}\hat{a}-\hat{b})}^2$

$={\left|\right|b_i-\stackrel{\‾}{b}\left|\right|}^2(1-{\left(\underset{i=1}{\overset{n}{\mathrm{\Σ}}}\hat{b}\hat{a}\right)}^2)---\left(2\right)$

Wherein for each sub-block determined,It is known, in order to obtain minimum match error RMS,Value require the smaller the better, in the matching process of each sub-block, pre-search restrictive condition is: 0.9 < m <1。

In described step 8, Block-matching being used matching error criterion, sub-block with matching error RMS of father's block is:

$\mathrm{RMS}=\frac{1}{N}[\underset{i=1}{\overset{N}{\mathrm{\&Sigma;}}}{r}_i^2+s(s\underset{i=1}{\overset{N}{\mathrm{\&Sigma;}}}{d}_i^2-2\underset{i=1}{\overset{N}{\mathrm{\&Sigma;}}}{r}_i{d}_i+2o\underset{i=1}{\overset{N}{\mathrm{\&Sigma;}}}{d}_i^2)+o(N\&CenterDot;o-2\underset{i=1}{\overset{N}{\mathrm{\&Sigma;}}}{r}_i)]---\left(3\right)$

Wherein parameter s and o are respectively as follows:

$s=\frac{[N\underset{i=1}{\overset{N}{\mathrm{\&Sigma;}}}{r}_i{d}_i-\underset{i=1}{\overset{N}{\mathrm{\&Sigma;}}}{r}_i\underset{i=1}{\overset{N}{\mathrm{\&Sigma;}}}{d}_i]}{[N\underset{i=1}{\overset{N}{\mathrm{\&Sigma;}}}{d}_i^2-{\left(\underset{i=1}{\overset{N}{\mathrm{\&Sigma;}}}{d}_i\right)}^2]}---\left(4\right)$

$o=\frac{1}{N}[\underset{i=1}{\overset{N}{\mathrm{\&Sigma;}}}{r}_i-s\underset{i=1}{\overset{N}{\mathrm{\&Sigma;}}}{d}_i]---\left(5\right)$

Wherein, N is sub-block and the number of father's block pixel, r_iFor the pixel value of sub-block, d_iPixel value for father's block.

Calculate current macro block-matching error RMS in reference frame, wherein r_iIt is the pixel value of sub-block, d_iIt it is father's block Pixel value；If RMS is less than threshold gamma set in advance, record IFS coefficient, IFS coefficient include match block displacement vector (x, Y) with formula (4), s and o in (5), process next macro block；Otherwise, current macro is carried out tree-shaped division, little after computation partition The RMS of block, if less than threshold gamma, then stops dividing, otherwise continues to divide, until sub-block reaches smallest blocks set in advance and is Only.

A kind of object-based Fast Fractal video decompression method, it is characterised in that comprise the steps of

Step I: first read in compression information, including compression frame number, the width of every two field picture and height, I frame compression quality is with slotting Enter the quality of I frame；

Step II: judge to decode whether frame is I frame, if I frame proceeds to step III, otherwise proceed to step IV；

Step III: for I frame, reads in code stream from compressed file, and the Alpha plane reading in this frame is decoded, and solves File after Ma includes video file based on different objects and complete video file, at object-based video file In, according to Alpha plane, the pixel belonging to this object retains, and is not belonging to the pixel zero setting of this object, and frame number adds one and proceeds to step Ⅵ；

Step IV: for non-I frame, first calculates in reference frame according to setting all macro blocks of step-length division and through tree-shaped Divide the pixel of fritter obtained and, pixel quadratic sum, from compressed file, then read in division information and the Huffman code of block Stream and the Alpha plane of this frame, thus obtain the dividing mode of the non-all macro blocks of I frame and the iterated function series of each fritter System coefficient, forwards step V to；Described reference frame is the most encoded former frame built of laying equal stress on of present frame；

Step V: use and remove square loop circuit filtering method: first the type on border is judged, defined parameters block edge Intensity, the most different, if frame for the block edge of varying strength, the wave filter of selection and the pixel number of required filtering Interior coding and be macroblock boundaries, then use strong filtering；If not intraframe coding and be not macroblock boundaries, affine block boundary uses one Level filtering, nonaffine block boundary need not filtering；Other situations use secondary filter；Finally it is decoded according to each macro block； Described affine piece is the block obtained by affine transformation, and described nonaffine block is not to be the block obtained by affine transformation；

Step VI: judge that the most all frames decode the most, if all decoding complete, terminating decoding process, otherwise proceeding to Step II.

When each macro block is decompressed, first determine whether this macro block dividing mode when coding, for each Individual sub-block, first finds the region corresponding with this sub-block in father's block territory, then utilizes equation below to obtain the picture of this sub-block Element value:

r_i=s d_i+o (6)

Wherein r_iFor the pixel value of sub-block to be decoded, d_iFor the pixel value in father's block territory, S is scale factor, O for skew because of Son.

During object-based decoding, the pixel only belonging to this subject area in current block is just decoded, with Sample, only utilizes the pixel belonging to same target region to be decoded in father's block territory, if in the middle part of certain sub-block in father's block territory Point pixel is not belonging to this object video, then the value of this partial pixel is with pixel average belonging to this subject area in this sub-block Value replaces.

Block edge intensity BS in described step V represents；Wherein, P₀',Q₀',P₁',Q₁' represent filtered pixel Value, P₀,P₁,Q₀,Q₁Represent that original pixel value, different BS and corresponding wave filter are as follows:

During BS=3, needing to filter by force, wave filter is expressed as:

P₀'=(P₁+P₀+Q₀)/3

Q₀'=(P₀+Q₀+Q₁)/3

P₁'=(2 P₁+P₀')/3 (7)

Q₁'=(2 Q₁+Q₀')/3

During BS=2, two-stage filter is expressed as:

P₀'=(P₁+2·P₀+Q₀)/4

Q₀'=(P₀+2·Q₀+Q₁)/4 (8)

During BS=1, one-level wave filter is expressed as:

P₀'=(P₁+3·P₀+Q₀)/5

Q₀'=(P₀+3·Q₀+Q₁)/5 (9)

As BS=0, it is not filtered.

The video sequence processed is yuv format, respectively at six steps above-mentioned to each employing in 3 components Reason.

One proposed by the invention object-based Fast Fractal video-frequency compression method advantage is:

(1) this method introduces fraction pixel block-matching technique in fractal coding algorithm, for a lot of blocks, in one It is inserted into the region of half-pixel accuracy and scans for finding preferably coupling, in order to obtain more accurate motion vector and more High compression ratio.

(2) this method is in the matching process of each sub-block, makes full use of fractal image feature, have employed pre-search and limits Condition, removes the father's block less mated in advance, improves coding efficiency and speed.

(3) this method is before carrying out the Block-matching of sub-block, calculates the macro block of present frame non-overlapping copies and through tree-shaped stroke The pixel of the fritter obtained after/and with pixel quadratic sum.In reference frame, according to coupling step-length calculate respectively each macro block with And the pixel of the fritter obtained after tree-shaped division and with pixel quadratic sum, calculate fractional pixel interpolation value correspondence fritter simultaneously Pixel and, pixel quadratic sum.Thus avoid the drawback duplicating calculating during Block-matching, be greatly saved son The match time of block.

(4) this method utilizes the asymmetric cross multi-level hexagon lattice point of improvement in the matching process of each sub-block Searching algorithm, fully combines characteristics of image and fractal compression, substantially increases coding rate.

(5) method introduces object-based video coding method, object-based coded method is being not based on object Further improve the performance of fractal compression on the basis of method, not only add compression ratio and Y-PSNR, and Compression is made to have greater flexibility.

(6) this method is when decoding, utilizes and removes square loop circuit filtering method, improves decoded image quality, for follow-up Subblock coding more preferably reference frame is provided.

Accompanying drawing explanation

Fig. 1 (a) is the compression process figure of a kind of object-based Fast Fractal video compress of the present invention and decompression method；

Fig. 1 (b) is the decompression flow process of a kind of object-based Fast Fractal video compress of the present invention and decompression method Figure；

Fig. 2 (a) is the 3rd frame of standard testing video sequence " mother-daughter.cif "；

Fig. 2 (b) is that a kind of object-based Fast Fractal video compress of the present invention is surveyed with the standard that decompression method obtains The Alpha segmentation plane of the 3rd frame of examination video sequence " mother-daughter.cif "；

Fig. 2 (c) is that a kind of object-based Fast Fractal video compress of the present invention individually decodes through this with decompression method The foreground video of the 3rd frame of the standard testing video sequence " mother-daughter.cif " that inventive method compressed encoding is later The result images of object；

Fig. 2 (d) is that a kind of object-based Fast Fractal video compress of the present invention individually decodes through this with decompression method The background video of the 3rd frame of the standard testing video sequence " mother-daughter.cif " that inventive method compressed encoding is later The result images of object；

Fig. 2 (e) is that a kind of object-based Fast Fractal video compress of the present invention individually decodes through this with decompression method 3rd frame result images of the standard testing video sequence " mother-daughter.cif " that inventive method compressed encoding is later；

Fig. 2 (f) is that a kind of object-based Fast Fractal video compress of the present invention individually decodes warp with decompression method 3rd frame result images of the standard testing video sequence " mother-daughter.cif " that CPM/NCIM method is later；

Fig. 3 (a) is in a kind of object-based Fast Fractal video compress of the present invention and decompression method UMHexagonS algorithm search route map；

Fig. 3 (b) is in a kind of object-based Fast Fractal video compress of the present invention and decompression method The starting point prognostic chart of UMHexagonS algorithm；

Fig. 4 (a) is a kind of object-based Fast Fractal video compress of the present invention and the fraction pixel in decompression method The half pixel interpolation schematic diagram of Block-matching；

Fig. 4 (b) is a kind of object-based Fast Fractal video compress of the present invention and the fraction pixel in decompression method The integral point fractional matching schematic diagram of Block-matching；

Fig. 5 (a) is a kind of object-based Fast Fractal video compress of the present invention and to macro block four kinds of decompression method Partition mode figure；

Fig. 5 (b) is a kind of object-based Fast Fractal video compress of the present invention and the decompression method division to macro block Pattern four carries out the four kinds of partition mode figures divided further；

Fig. 6 (a) be a kind of object-based Fast Fractal video compress of the present invention with decompression method to object-based The labelling figure of three kinds of image blocks；

Fig. 6 (b) is that boundary block is belonged to by a kind of object-based Fast Fractal video compress of the present invention with decompression method The labelling figure of the pixel of different video subject area；

Fig. 7 (a) is a kind of object-based Fast Fractal video compress of the present invention and remove square ring in decompression method The vertical boundary of road filtering faces the sampling schematic diagram in territory；

Fig. 7 (b) is a kind of object-based Fast Fractal video compress of the present invention and remove square ring in decompression method The block edge intensity decision tree schematic diagram of road filtering；

Fig. 8 (a) is a kind of object-based Fast Fractal video compress of the present invention and decompression method and tradition CPM/ NCIM method is to the Y-PSNR that 3～9 frames of standard testing video sequence " mother-daughter.cif " are compressed Comparison diagram；

Fig. 8 (b) is a kind of object-based Fast Fractal video compress of the present invention and decompression method and tradition CPM/ The contrast to the compression ratio that 3～9 frames of standard testing video sequence " mother-daughter.cif " are compressed of the NCIM method Figure；

Fig. 8 (c) is a kind of object-based Fast Fractal video compress of the present invention and decompression method and tradition CPM/ NCIM method is right to the compression time that 3～9 frames of standard testing video sequence " mother-daughter.cif " are compressed Than figure.

Detailed description of the invention

Being described in further detail the inventive method below in conjunction with accompanying drawing, only as a example by luminance component Y, aberration divides The compression step of amount U with V is identical with luminance component.

The present invention proposes a kind of object-based Fast Fractal video compress and decompression method, divides first with video Segmentation method obtains video object segmentation plane i.e. Alpha plane, start frame uses block dct transform coding, non-I frame is carried out block Motion estimation/compensation encodes, calculate to the pixel of the relevant sub-block in sub-block territory and father's block territory and with pixel quadratic sum, in fraction pixel The pixel of interpolation corresponding blocks and, pixel quadratic sum, carry out pre-search restrictive condition judgement, and utilize in former frame search window point Number block of pixels coupling and the asymmetric cross multi-level hexagonal point search algorithm improved find most like match block, profit With Huffman compression coding iterated function system coefficient.Homographic solution compression process is: non-I frame carries out Huffman decoding and obtains Iterated function system coefficient, carries out decoding based on macro block, calculate father's block territory be correlated with sub-block pixel and with pixel quadratic sum, so After successively each macro block in present frame is decoded, and utilize remove square loop circuit filtering method.

As shown in accompanying drawing 1 (a), a kind of object-based Fast Fractal video-frequency compression method, comprise the following steps:

Step one: as a example by front 10 frames of standard testing video sequence " mother-daughter.cif ".Utilize and automatically regard Frequently video sequence is split by object segmentation methods automatically, obtains Alpha plane.Object in video can pass through Alpha Plane is defined, and can independently be compressed coding.Accompanying drawing 2 (a) is standard testing video sequence " mother- Daughter.cif " the 3rd frame, accompanying drawing 2 (b) is the Alpha plane of this frame, and prospect white represents, background black represents. As shown in accompanying drawing 2 (a), image has two object videos, then Alpha plane can be just a bianry image, by white Representing prospect, black represents background.Each object can be compressed independently so that each object one code stream of composition. So when decompressing, it is not necessary to obtain entire image, but can individually recover and control each object.Before individually decompressing The result of scape object video is accompanying drawing 2 (c), and the result individually decompressing background video object is accompanying drawing 2 (d), and whole frame decompresses Result be accompanying drawing 2 (e), utilizing the decompressing image after CPM/NCIM is accompanying drawing 2 (f).Introduce object-based coding not But improve compression performance, and add the motility of method for video coding.

Step 2: to video sequence " mother-daughter.cif " start frame, it is first determined whether be I frame, if I frame, uses I frame intraimage method based on block dct transform, start frame is divided into the son of the non-overlapping copies of 8 × 8 Block, carries out dct transform respectively to each sub-block.Described I frame is only to carry out in frame in video sequence start frame or video sequence The picture frame of coding；Block in described piece of dct transform uses fixed size pattern.Discrete cosine transform by 8 × 8 image pattern X, is transformed into the coefficient matrix Y of 8 × 8.Conversion process (including inverse transformation) can represent with transformation matrix A.

The forward DCT(FDCT of 8 × 8 sample block) conversion as follows:

Y=AXA^T (10)

Reversely DCT(IDCT) as follows:

X=A^TYA (11)

Wherein A is the transformation matrix of 8 × 8.Each element in A is as follows:

$A_{\mathrm{ij}}=C_i\cos \frac{(2j+1)\mathrm{i\&pi;}}{16}---\left(12\right)$

Wherein

$C_i=\sqrt{\frac{1}{8}}(i=0)$ $C_i=\frac{1}{2}(i>0)---\left(13\right)$

I, j are respectively the row and column of matrix A.

Conversion coefficient quantified and encodes, proceeding to step 10；Otherwise, step 3 is forwarded to；

Step 3: 16 × 16 macro blocks of the non-overlapping copies that current non-I two field picture is fixed size divide, and calculate respectively The 16 × 16 of present frame non-overlapping copies, the pixel of the fritter of 16 × 8,8 × 16,8 × 8,8 × 4,4 × 8,4 × 4 and with pixel Quadratic sum.Calculate the most respectively in the decompressing image of reference frame i.e. former frame the size with 1 as step-length be respectively 16 × The pixel of the fritter of 16,16 × 8,8 × 16,8 × 8,8 × 4,4 × 8,4 × 4 and the quadratic sum with pixel, calculate mark picture simultaneously Element interpolated value correspondence 16 × 16,16 × 8,8 × 16,8 × 8,8 × 4,4 × 8,4 × 4 fritter pixel and, pixel quadratic sum, With the double counting during minimizing Block-matching.Start to process from first macro block, successively all 16 × 16 grand to present frame Block encodes, and proceeds to step 4；The image block of the described non-overlapping copies that present frame is divided into fixed size is referred to as macro block；Institute State and current macro is carried out the block that tree-shaped division obtains be referred to as fritter；Described present frame is the frame being compressed；Described ginseng Examine the most encoded former frame built of laying equal stress on that frame is present frame；The collection that described present frame is all pieces is collectively referred to as sub-block territory；Before described The collection of all pieces of one frame is collectively referred to as father's block territory；

Step 4: in the current frame according to non-overlapping copies 16 × 16 macro block carry out Block-matching, according to the class of this macro block Type, makes following selection, if this macro block is not in Alpha plane, does not processes this block, continues with next macro block； If this macro block is entirely located in Alpha plane, proceed to step 5；If this macro-block partitions is positioned at Alpha plane, proceed to step Rapid nine.If after present frame macro block all mates, proceeding to step 10.Described not in the object video region of present encoding Block be referred to as external block, described in be all within the object video region of present encoding block be referred to as internal block, described part picture The element not block in the object video region of present encoding is boundary block.

Step 5: successively all 16 × 16 macro blocks of present frame are encoded, to whole in the search window in father's block territory Individual macro block carries out block motion estimation/compensation；In carrying out the sub-block matching process with father's block, the position of sub-block rising as father's block Beginning Searching point, the size of father's block is identical with the size of sub-block, forwards step 6 to；

Step 6: utilize the asymmetric cross multi-level hexagonal point search algorithm and fraction pixel block improved Join, search out optimal matching error.The asymmetric cross multi-level hexagonal point search algorithm improved, its initial search Shown in route such as Fig. 3 (a), compared to the asymmetrical hexagonal-shaped algorithm in H.264, the improvement of this method is mainly reflected in following three Point:

1) starting point prediction:

It is not involved with multi-reference frame based on fractal video coding algorithm, and macro block and sub-block have different big Little, therefore three kinds of modes as shown in Fig. 3 (b) of utilization carry out starting point prediction:

A) spatial domain median prediction: take a left side for current sub-block, upper, the motion vector intermediate value of right adjacent block is that predicted motion is vowed Amount；

B) initial point prediction: according to temporal correlation, makes motion vector value for (0,0)；

C) neighboring reference frame prediction: utilize the MV of correspondence position block in previous reference frame to be predicted in proportion；

2) threshold value jump condition during asymmetric cross template search:

The sub-block of fractal image and error matching criterior R of father's block are formula (3), (4), (5).The choosing that varies in size according to block Selecting different threshold values, asymmetric cross template search is complete, selects optimal match point and carries out follow-up masterplate as new starting point Coupling；

3) end condition is shifted to an earlier date:

Feature according to fractal coding algorithm will terminate being divided into two kinds of situations in advance: one is at non-homogeneous multi-level hexagon In lattice point whole pixel motion search procedure, in addition to the end condition in advance of this algorithm itself, in order to reduce search complexity such as Really optimum point is positioned at hexagonal centre, can stop search；Two is to use tree-shaped partition structure based on fractal video coding algorithm. As shown in Fig. 5 (a), it is first according to pattern 1 and carries out non-uniform multilayer hexaploid pixel motion searching method, if full Foot threshold condition, then terminate the coding of this macro block, carry out the coding of next macro block, otherwise according to pattern 2 by the macro block of pattern 1 Divide, each sub-block is carried out the estimation of non-uniform multilayer hexaploid pixel motion searching method, and By that analogy.

Then the RMS point at search fractional pixel interpolation value correspondence fritter, step is as follows:

1) pixel in region of search in reference frame is carried out interpolation and forms one relative to the pixel in integer position more High-resolution region；

2) carry out integer pixel in interpolation region and optimal coupling is found in half-pixel position search；

3) current block is substituted by the affine transformation of match block.

As shown in Fig. 4 (a), a represents original integer pixel, after b and c represents by a pair integer pixel a linear interpolation Pixel value, d represent by the pixel value after around four integer pixel a linear interpolations, arrow represents interior direction interpolation.Until looking for To minimum RMS point, as shown in Fig. 4 (b), it is assumed that A point is integer-pixel search optimum point, carrying out fraction pixel about Motion search, such as point 1,2,3,4,5,6,7,8, although amount of calculation increased, but half pel motion is estimated and motion compensation Performance significantly better than the Motion estimation and compensation of integer pixel, forward step 7 to；

Step 7: pre-search restrictive condition judges: for specific sub-block, has following derivation, wherein, b_iPicture for sub-block Element value, a_iFor the pixel value of father's block, s is the scale factor in fractal image, and o is displacement factor, and | | a | | represents two dimension norm, That is:

| | a | |=(| a₁|²+|a₂|²+···+|a_n|²)^1/2:

$\mathrm{RMS}=\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}{(s\&CenterDot;a_i+o-b_i)}^2$

$=\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}{(s\&CenterDot;a_i+\frac{1}{n}[\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}{b}_i-s\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}{a}_i]-b_i)}^2$

$=\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}{((a_i-\frac{\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}{a}_i}{n})\&CenterDot;\frac{[n\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}{a}_i{b}_i-\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}{a}_i\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}{b}_i]}{[n\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}{a}_i^2-{\left(\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}{a}_i\right)}^2]}+\frac{\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}{b}_i}{n}-b_i)}^2$

$=\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}{((a_i-\stackrel{\&OverBar;}{a})\&CenterDot;\frac{[\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}{a}_i{b}_i-n\stackrel{\&OverBar;}{a}\stackrel{\&OverBar;}{b}]}{[\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}{a}_i^2-n{\stackrel{\&OverBar;}{a}}^2]}+\stackrel{\&OverBar;}{b}-b_i)}^2$

$=\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}{((a_i-\stackrel{\&OverBar;}{a})\&CenterDot;\frac{\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}(b_i-\stackrel{\&OverBar;}{b})(a_i-\stackrel{\&OverBar;}{a})}{{\left|\right|a_i-\stackrel{\&OverBar;}{a}\left|\right|}^2}+\stackrel{\&OverBar;}{b}-b_i)}^2$

$={\left|\right|b_i-\stackrel{\&OverBar;}{b}\left|\right|}^2\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}{(\frac{(a_i-\stackrel{\&OverBar;}{a})}{\left|\right|a_i-\stackrel{\&OverBar;}{a}\left|\right|}\&CenterDot;\frac{\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}(b_i-\stackrel{\&OverBar;}{b})(a_i-\stackrel{\&OverBar;}{a})}{\left|\right|b_i-\stackrel{\&OverBar;}{b}\left|\right|\left|\right|a_i-\stackrel{\&OverBar;}{a}\left|\right|}-\frac{b_i-\stackrel{\&OverBar;}{b}}{\left|\right|b_i-\stackrel{\&OverBar;}{b}\left|\right|})}^2---\left(14\right)$

Allow $\hat{a}=\frac{(a_i-\stackrel{\&OverBar;}{a})}{\left|\right|a_i-\stackrel{\&OverBar;}{a}\left|\right|},$ $\hat{b}=\frac{b_i-\stackrel{\&OverBar;}{b}}{\left|\right|b_i-\stackrel{\&OverBar;}{b}\left|\right|},$ And understand ${\left|\right|\hat{a}\left|\right|}^2=1,$ ${\left|\right|\hat{b}\left|\right|}^2=1,$ Then R can derive as follows:

$\mathrm{RMS}={\left|\right|b_i-\stackrel{\&OverBar;}{b}\left|\right|}^2\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}{(\hat{a}\&CenterDot;\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}\hat{b}\hat{a}-\hat{b})}^2$

$={\left|\right|b_i-\stackrel{\&OverBar;}{b}\left|\right|}^2(1-{\left(\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}\hat{b}\hat{a}\right)}^2)---\left(15\right)$

Wherein for each sub-block determined,It is known, therefore to obtain minimum match error RMS；Value require the smaller the better, in the matching process of each sub-block, pre-search restrictive condition is: 0.9<m<1.If meeting pre-search restrictive condition with father's block respective value, then forward step 8 to；The most directly preserve current iteration Function system coefficient i.e. IFS coefficient, proceeds to step 4 and encodes next macro block；

Step 8: tree-shaped division is mated further: the foundation of coupling is fractal iteration function system principle, briefly introduces one The Fundamentals of Mathematics iterated function system (IFS:Iterative Function System) of lower Fractal Image Compression is theoretical.If D It is RⁿThe subset of Euclidean space, ω is the mapping of D → D, if there is a real number C, 0≤C ＜ 1 so that for RⁿOn tolerance D, meets any x, y ∈ D, and (d (x, y)), then ω is called that compression maps, and real number C is referred to as ω's d (ω (x), ω (y))≤C Compressibility factor.Complete metric space (X, d) and n compression maps ω_i: its compressibility factor of X → X(is respectively C₁, C₂,···C_n) together, just one iterated function system (Iterated Function System) of composition, it is called for short IFS, is denoted as {X:ω₁,ω₂,···,ω_n}.C=max (C₁,C₂,···,C_n) be referred to as IFS compressibility factor.Therefore { R²:ω₁,ω₂, ω₃It is exactly an IFS.

In Fractal Image Compression, general matching criterior is RMS, it may be assumed that

$\mathrm{RMS}=\frac{1}{N}[\underset{i=1}{\overset{N}{\mathrm{\&Sigma;}}}{r}_i^2+s(s\underset{i=1}{\overset{N}{\mathrm{\&Sigma;}}}{d}_i^2-2\underset{i=1}{\overset{N}{\mathrm{\&Sigma;}}}{r}_i{d}_i+2o\underset{i=1}{\overset{N}{\mathrm{\&Sigma;}}}{d}_i^2)+o(N\&CenterDot;o-2\underset{i=1}{\overset{N}{\mathrm{\&Sigma;}}}{r}_i)]---\left(16\right)$

Wherein s, o are respectively as follows:

$s=\frac{[N\underset{i=1}{\overset{N}{\mathrm{\&Sigma;}}}{r}_i{d}_i-\underset{i=1}{\overset{N}{\mathrm{\&Sigma;}}}{r}_i\underset{i=1}{\overset{N}{\mathrm{\&Sigma;}}}{d}_i]}{[N\underset{i=1}{\overset{N}{\mathrm{\&Sigma;}}}{d}_i^2-{\left(\underset{i=1}{\overset{N}{\mathrm{\&Sigma;}}}{d}_i\right)}^2]}---\left(17\right)$

$o=\frac{1}{N}[\underset{i=1}{\overset{N}{\mathrm{\&Sigma;}}}{r}_i-s\underset{i=1}{\overset{N}{\mathrm{\&Sigma;}}}{d}_i]---\left(18\right)$

Wherein, N is sub-block and the number of father's block pixel, r_iFor the pixel value of sub-block, d_iPixel value for father's block.

First setting the match error threshold γ=tol × tol × no of sub-block, wherein tol is according to different sub-block sizes Changing, big sub-block tol is the biggest, and little sub-block tol is the least.In this example, we take the tol of 16 × 16 macro blocks is 10.0,8 The tol of × 8 sub-blocks be the tol of 8.0,4 × 4 sub-blocks be 6.0, no is the pixel that current sub-block belongs to this object video region Number.

First match error threshold γ of 16 × 16 macro blocks is set₁₆=10.0 × 10.0 × no, in father's block territory of reference frame Start in the search window of 15 × 15, whole macro block to be carried out Block-matching, if matching error RMS is little with the position of current sub-block In the threshold gamma starting setting₁₆, then preserving current IFS coefficient and include scale factor s, offset o, father's block is relative to current sub-block Coordinate offset x, y, return step 4, continue the coupling of next macro block.

Otherwise, dividing this macro block according to tree, the division to macro block has four kinds of patterns, such as accompanying drawing 5 (a), Pattern one is 16 × 16 fritters, and pattern two is the fritter of two 8 × 16, and pattern three is the fritter of two 16 × 8, pattern four It it is the fritter of four 8 × 8.

1, first calculate by the division of pattern two, utilize the asymmetric cross multi-level hexagonal point search improved to calculate Method and fraction pixel Block-matching, if two fritters all meet RMS < γ in pattern two₁₆, then preserve current IFS coefficient and include ratio Example factor s, offsets o, and father's block is relative to coordinate offset x, the y of current sub-block, and the division of stop piece, forwards 5 to；

2, otherwise divide by pattern three, utilize the asymmetric cross multi-level hexagonal point search algorithm improved and divide Number block of pixels coupling, if two fritters all meet RMS < γ in pattern three₁₆, then preserve current IFS coefficient and include scale factor S, offsets o, and father's block is relative to coordinate offset x, the y of current sub-block, and the division of stop piece, forwards 5 to；

3, otherwise according to pattern four, current macro is divided, utilize the multi-level hexagon of asymmetric cross improved Grid search algorithm and fraction pixel Block-matching, now match error threshold is set to γ₈=8.0 × 8.0 × no, if pattern 4 fritters in four all meet RMS < γ₈, then preserve current IFS coefficient and include scale factor s, offset o, and father's block phase For coordinate offset x, the y of current sub-block, and the division of stop piece, forward 5 to；

4, otherwise each fritter in pattern four is divided according to the mode division order in accompanying drawing 5 (b), can depend on The secondary fritter being divided into 18 × 8, the fritter of 24 × 8, the fritter of 28 × 4, the fritter of 44 × 4.Here only to first The matching process of individual 8 × 8 fritters is illustrated, the matching process of other 38 × 8 fritters and first identical, repeat no more. It is first according to the fritter division of 24 × 8, carries out Block-matching, if matching error RMS of two sub-blocks is all less than γ₈Time, Then preserve current IFS coefficient and include scale factor s, offset o, and father's block is relative to coordinate offset x, the y of current sub-block, and The division of stop piece.Otherwise, carry out the division of block according to the dividing mode of 28 × 4, the two sub-block is carried out Block-matching, as Really matching error RMS of two sub-blocks is all less than γ₈Time, then preserve current IFS coefficient and include scale factor s, offset o, And father's block is relative to coordinate offset x, the y of current sub-block, and the division of stop piece.Otherwise, it it is 44 × 4 to this partition Fritter, match error threshold is set to γ simultaneously₄Four fritters are carried out Block-matching, and remember respectively by=6.0 × 6.0 × no respectively The IFS coefficient recording each sub-block includes scale factor s, offsets o, and father's block is relative to coordinate offset x, the y of current sub-block, and The division of stop piece, forwards 5 to；

5, return step 4, continue the coding of next macro block.

The most encoded complete if all of macro block, then forward step 10 to；

Step 9: in order to when matched sub-block is with father's block, it is to avoid the pixel belonging to different object is obscured, mutually to each Pixel does a labelling in Alpha plane, indicates which object is labeled pixel be belonging to, as shown in accompanying drawing 6 (a), The pixel of this boundary block is marked as two parts of S1 and S2.

The concrete compression method of boundary block: assume current compression be object 1(compressed object 2 time, method is identical), i.e. S1 The object at place.For the amount relevant to sub-block, only calculate the pixel value in S1 region, and the pixel in S2 region does not gives Consider；For the amount relevant with father's block, if a certain pixel d in father's block of the position corresponding with sub-block_iFall within S1 region, Then use d_iOriginal pixel value, otherwise, replace d according to specific value of calculation_i, the present invention uses in father's block and belongs to S1 The pixel average in region replaces d_i.Coupling maps shown in effect such as accompanying drawing 6 (b).It should be noted that coupling maps only same Carry out between class block, i.e. sub-block and father's block must be boundary block simultaneously or be internal block (external block) simultaneously, returns step 5 Process；

Step 10: all IFS coefficients carry out Huffman coding, Huffman coding is by each according to the probability occurred Symbol is mapped in the set (VLC) of an avriable length codes, reduces the statistical redundancy of IFS coefficient data.Whether judge present frame For last frame, if last frame terminates coding, otherwise, proceed to step 2 and continue with next frame image.

As shown in figure ib, a kind of object-based Fast Fractal video decompression method, comprise the following steps:

Step I: first read in compression information, including compression frame number, the width of every frame and height, I frame reconstruction quality, inserts I frame Interval etc.；

Step II: judge to decode whether frame is I frame, if I frame proceeds to step III, otherwise proceed to step IV；

Step III: for I frame, reads in code stream from compressed file, and the Alpha plane reading in this frame is decoded, and enters Row inverse DCT converts, and obtains the pixel value of each block of 8 × 8, and the file after decoding includes video literary composition based on different objects Part and complete video file, in object-based video file, according to Alpha plane, the pixel belonging to this object retains, Being not belonging to the pixel zero setting of this object, frame number adds one and proceeds to step VI；

Step IV: for non-I frame, first calculates in reference frame according to setting all macro blocks of step-length division and through tree-shaped Divide the pixel of fritter obtained and, pixel quadratic sum, from compressed file, then read in division information and the Huffman code of block Stream and the Alpha plane of this frame, thus obtain the dividing mode of the non-all macro blocks of I frame and the iterated function series of each fritter System coefficient, is decoded according to each macro block, when decompressing for each macro block, first determines whether that this macro block is when coding Dividing mode, for each sub-block, first find the region corresponding with this sub-block in father's block territory, then utilize following Formula obtains the pixel value of this sub-block:

r_i=s d_i+o

Wherein r_iFor the pixel value of sub-block to be decoded, d_iFor the pixel value in father's block territory, S is scale factor, O for skew because of Son.

During object-based decoding, the pixel only belonging to this subject area in current block is just decoded, with Sample, only utilizes the pixel belonging to same target region to be decoded in father's block territory, if in the middle part of certain sub-block in father's block territory Point pixel is not belonging to this object video, then the value of this partial pixel is with pixel average belonging to this subject area in this sub-block Value replaces, and described reference frame is the most encoded former frame built of laying equal stress on of present frame, proceeds to step V；

Step V: use and remove square loop circuit filtering method: first the type on border is judged, defined parameters block edge Intensity, the most different for the block edge of varying strength, the wave filter of selection and the pixel number of required filtering, such as Fig. 7 A shown in (), vertical boundary faces the sampling schematic diagram in territory, if intraframe coding and be macroblock boundaries, then use and filter by force；If no Being intraframe coding and be not macroblock boundaries, affine block boundary uses one-level filtering, and nonaffine block boundary need not filtering；Other feelings Condition uses secondary filter；Finally it is decoded according to each macro block；Specifically chosen as shown in Fig. 7 (b), block edge intensity BS table Show, wherein, P₀',Q₀',P₁',Q₁' represent filtered pixel value, P₀,P₁,Q₀,Q₁Represent original pixel value, different BS and Corresponding wave filter is as follows, and described affine piece is the block obtained by affine transformation, and described nonaffine block is not is by affine transformation The block obtained:

During BS=3, needing to filter by force, wave filter is expressed as:

P₀'=(P₁+P₀+Q₀)/3

Q₀'=(P₀+Q₀+Q₁)/3 (19)

P₁'=(2 P₁+P₀')/3

Q₁'=(2 Q₁+Q₀')/3

During BS=2, two-stage filter is expressed as:

P₀'=(P₁+2·P₀+Q₀)/4 (20)

Q₀'=(P₀+2·Q₀+Q₁)/4

During BS=1, one-level wave filter is expressed as:

P₀'=(P₁+3·P₀+Q₀)/5 (21)

Q₀'=(P₀+3·Q₀+Q₁)/5

As BS=0, it is not filtered.

It is decoded according to each macro block, when each macro block is decompressed, first determines whether that this macro block is at coding Time dividing mode, for each sub-block, first find the region corresponding with this sub-block in father's block territory, then utilize following Formula obtain the pixel value of this sub-block:

r_i=s d_i+o (22)

Wherein r_iFor the pixel value of sub-block to be decoded, d_iFor the pixel value in father's block territory, S is scale factor, O for skew because of Son.

Step VI: judge that the most all frames decode the most, if all decoding complete, terminating decoding process, otherwise proceeding to Step II.

The video sequence processed is yuv format, respectively at six steps above-mentioned to each employing in 3 components Reason.

This method selects visual c++ 6.0 as the language that realizes of described method, and CPU is Core^TM2DuoT8300,2.4GHz dominant frequency, memory size is 2G, to standard testing video sequence " mother- Daughter.cif " carry out quick fractal video coding experiments.

The compression performance average contrast of table 1 video sequence

It is respectively adopted traditional CPM/NCIM method and the inventive method 3～9 frames to " mother-daughter.cif " Shown in the comparison diagram such as accompanying drawing 8 (a) of the Y-PSNR being compressed coding；It is respectively adopted traditional CPM/NCIM method and basis Inventive method is compressed the comparison diagram such as accompanying drawing 8 of the compression ratio of coding to front 3～9 frames of " mother-daughter.cif " Shown in (b)；It is respectively adopted traditional CPM/NCIM method and the inventive method front 3～9 to " mother-daughter.cif " Shown in the comparison diagram of the time that frame is compressed such as accompanying drawing 8 (c)；From accompanying drawing 8 and table 1 it can be seen that the inventive method is with traditional CPM/NCIM method compare, not only compression time is kept to 0.087 times, and Y-PSNR PSNR adds about 8dB, pressure Contracting ratio adds about about 120.

Claims (4)

1. an object-based Fast Fractal video decompression method, it is characterised in that comprise the steps of

Step I: first read in compression information, including compression frame number, the width of every two field picture and height, I frame compression quality and insertion I frame Quality；

Step II: judge to decode whether frame is I frame, if I frame proceeds to step III, otherwise proceed to step IV；

Step III: for I frame, reads in code stream from compressed file, and the Alpha plane reading in this frame is decoded, and decodes it After file include video file based on different objects and complete video file, in object-based video file, root According to Alpha plane, the pixel belonging to this object retains, and is not belonging to the pixel zero setting of this object, and frame number adds one and proceeds to step VI；

Step IV: for non-I frame, first calculates in reference frame according to setting all macro blocks of step-length division and through tree-shaped division The pixel of the fritter obtained and, pixel quadratic sum, then read in from compressed file the division information of block and Huffman code stream with And the Alpha plane of this frame, thus obtain the dividing mode of the non-all macro blocks of I frame and the iterated function system system of each fritter Number, forwards step V to；Described reference frame is the most encoded former frame built of laying equal stress on of present frame；

Step V: use and remove square loop circuit filtering method: first judging the type on border, defined parameters block edge is strong Degree, the most different, if in frame for the block edge of varying strength, the wave filter of selection and the pixel number of required filtering Coding and be macroblock boundaries, then use and filter by force；If not intraframe coding and be not macroblock boundaries, affine block boundary uses one-level Filtering, nonaffine block boundary need not filtering；Other situations use secondary filter；Finally it is decoded according to each macro block；Institute Stating affine piece is the block obtained by affine transformation, and described nonaffine block is not to be the block obtained by affine transformation；

Step VI: judge that the most all frames decode the most, if all decoding complete, terminating decoding process, otherwise proceeding to step Ⅱ；

Block edge intensity BS in described step V represents；Wherein, P₀',Q₀',P₁',Q₁' represent filtered pixel value, P₀, P₁,Q₀,Q₁Representing original pixel value, its distributing order is followed successively by P from left to right₁, P₀, Q₀, Q₁, different BS and corresponding filter Ripple device is as follows:

During BS=3, needing to filter by force, wave filter is expressed as:

$\begin{array}{c}{P_0}^{\&prime;}=(P_1+P_0+Q_0)/3\\ {Q_0}^{\&prime;}=\left(P_0+Q_0+Q_1\right)/3\\ {P_1}^{\&prime;}=\left(2\&CenterDot;P_1+{P_0}^{\&prime;}\right)/3\\ {Q_1}^{\&prime;}=\left(2\&CenterDot;Q_1+{Q_0}^{\&prime;}\right)/3\end{array}---\left(2\right)$

During BS=2, two-stage filter is expressed as:

$\begin{array}{c}{P_0}^{\&prime;}=(P_1+2\&CenterDot;P_0+Q_0)/4\\ {Q_0}^{\&prime;}=(P_0+2\&CenterDot;Q_0+Q_1)/4\end{array}---\left(3\right)$

During BS=1, one-level wave filter is expressed as:

$\begin{array}{c}{P_0}^{\&prime;}=(P_1+3\&CenterDot;P_0+Q_0)/5\\ {Q_0}^{\&prime;}=(P_0+3\&CenterDot;Q_0+Q_1)/5\end{array}---\left(4\right)$

As BS=0, it is not filtered.

One the most according to claim 1 object-based Fast Fractal video decompression method, it is characterised in that: for When each macro block decompresses, first determine whether this macro block dividing mode when coding, for each sub-block, first exist The region corresponding with this sub-block is found in father's block territory, then utilizes equation below to obtain the pixel value of this sub-block:

r_i=s d_i+o (1)

Wherein r_iFor the pixel value of sub-block to be decoded, d_iFor the pixel value in father's block territory, S is scale factor, and O is displacement factor.

One the most according to claim 1 object-based Fast Fractal video decompression method, it is characterised in that: at base During the decoding of object, the pixel only belonging to this subject area in current block is just decoded, the most sharp in father's block territory It is decoded by the pixel belonging to same target region, if partial pixel is not belonging to this video in certain sub-block in father's block territory Object, then the value of this partial pixel meansigma methods of the pixel belonging to this subject area in this sub-block replaces.

One the most according to claim 1 object-based Fast Fractal video decompression method, it is characterised in that: process Video sequence be yuv format, six steps above-mentioned to each employing in 3 components process respectively.