CN103339952B - Content-adaptive compensation for loss and damage for efficient video coding filters - Google Patents

Content-adaptive compensation for loss and damage for efficient video coding filters Download PDF

Info

Publication number
CN103339952B
CN103339952B CN201180061488.3A CN201180061488A CN103339952B CN 103339952 B CN103339952 B CN 103339952B CN 201180061488 A CN201180061488 A CN 201180061488A CN 103339952 B CN103339952 B CN 103339952B
Authority
CN
China
Prior art keywords
loss
compensation
damage
video
frame
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Active
Application number
CN201180061488.3A
Other languages
Chinese (zh)
Other versions
CN103339952A (en
Inventor
A·普里
N·戈卡莱
D·索塞卡
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Intel Corp
Original Assignee
Intel Corp
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Priority claimed from US12/974,645 external-priority patent/US9462280B2/en
Application filed by Intel Corp filed Critical Intel Corp
Priority to CN201610854915.8A priority Critical patent/CN107071484B/en
Publication of CN103339952A publication Critical patent/CN103339952A/en
Application granted granted Critical
Publication of CN103339952B publication Critical patent/CN103339952B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Abstract

Describing the system and method for compensation for loss and damage filtering, it can be used in combination with Video coding, or be individually used for post processing.It combines efficient code book and represents use Wiener filtering method.

Description

Content-adaptive compensation for loss and damage for efficient video coding filters
Technical field
Present disclosure relates to coding and the decoding of video signal, and more particularly, to for compensation for loss and damage Sef-adapting filter carries out coding based on code book and decoding.
Background technology
The such as appearance of the digital multimedia of digital picture, voice/audio, figure and video etc has dramatically improved Various application, and due to relatively easy and open brand-new application, have been realized in whereby reliable memory to content, Communicate, transmit and search for and access.Generally speaking, it is enough in the application of digital multimedia, contains and includes amusement, letter Breath, medical treatment and safety are in interior broad range, and make society be benefited the most in many ways.Such as photographing unit and mike Etc the multimedia that captured of sensor be typically simulation, and the digitized process of pulse code modulation (PCM) form Become numeral.But, the most after digitization, the amount of produced data is probably sizable, it is necessary to again Create the analog representation required for speaker and/or TV display.Thus, the efficient biography to substantial amounts of digital multimedia content Send, store or transmit needs and be compressed to compression expression from original PCM form.Thus, invent many multimedia pressures Contracting technology.For many years, video compression technology has developed the most most advanced and sophisticated, has arrived at such degree: they usually can Enough realize the high compression rate between 10 to 100, keep high psycho visual quality, usually with unpressed digital video class simultaneously Seemingly.
Although the most having achieved greatly progress in field of video compression and science (as crossed multiple standard body Drive video encoding standard (such as MPEG-1, MPEG-2, H.263, MPEG-4part 2, MPEG-4AVC/H.264, MPEG- 4SVC and MVC) and the propriety standard (such as Windows Media Video, Real Video, On2VP etc.) of industry driving Represented), but consumer is (vertical for more high-quality, more fine definition and the present 3D being available for whenever and wherever possible accessing Body) appetite improved constantly of video has made (such as DVD/BD, air broadcast, cable/satellite, is had by various modes Line and mobile network) it is delivered to a series of client device (such as PC/ kneetop computer, TV, Set Top Box, game console, just Take formula media player/equipment and smart phone) necessitate, this has lighted the phase to the most high level video compress Hope.Standard body drive standard in, this by ISO MPEG in high efficiency Video coding recently effort proved, Expect that this effort is by new technical contribution and the exploration made H.265 video compress for many years from ITU-T standard committee The technical combinations of sex work is got up.
All aforesaid standards all use general inter prediction encoding framework, and it relates to reducing in the following manner the time Redundancy: by first dividing the frame into block and to each piece of distribution motion vector of frames to be encoded, to solve relative to the past Motion between frame of video is compensated by the frame of code;These motion vectors are subsequently sent to decoder and for generating and mistake Go the motion-compensated predicted frames that the frame of decoding is different, and generally encoded block by block by transform coding.Recognize Arrive: with the high compression rate of current decoder-coder operation today, motion compensation transform coding process may generate from coding noise (such as mosquito effect), block (blockiness), bleed into the coding artifact of ring (ringing).It practice, in order to alleviate this Some in a little coding artifact, most popular codec includes deblocking filter and is ring sometimes, to filter as post processing Ripple device.Compared with the post processing that can carry out outside loop, deblock and carry out in loop.The most such deblocking can be pressed down Visual quality artifact processed, but the infringement not to codec is introduced compensates, and such deblocking can not be given too Many coding quality gains.
Recognize this point, at several years of the past, have been carried out research make great efforts to be designed to improve visual quality and The additional filter of coding quality.Additionally, due to coding quality may be according to the content of video sequence, contrast ratio, noise grade And operate bit rate and change very big, it is desirable to wave filter is adaptive rather than fixing, to obtain such filter The maximum benefit of ripple device, thus improve prediction and thus help improve compression efficiency.For calculating the mathematics mistake of filter coefficient Journey is based on Wiener Hopf equation, and calculates filter coefficient (such as, the coefficient arrays of 5 × 5,7 × 7 or 9 × 9), for often For each frame of individual video sequence, directly or differentially send described filter coefficient.
According to the ultimate principle of adaptive-filtering, cross-correlation matrix and autocorrelation matrix are accumulated, by with lower section Wiener Hofp equation is solved by formula, can calculate optimum Wiener filtering from cross-correlation matrix and autocorrelation matrix Device.
Making x (n) is input signal, and y (n) is output, and h (n) represents filter coefficient.
Wave filter exports:
Error signal:
E (n)=d (n)-y (n)
Mean square error:
J=E [e2(n)]
Matrix form is:
And
Y (n)=hTX (n)=x (n)Th
Wherein, PdIt it is scalar.
Cross-correlation row vector:
Rdx=E [d (n) x (n)T]
Autocorrelation matrix:
Rxx=E [X (n) x (n)T]
In order to find minima, differentiate, and derivative is arranged to zero, as follows:
Solving for h, Wiener Hofpf equation is as follows:
Wiener Hofpf equation determines filter coefficient optimum in terms of mean square error, and the wave filter quilt produced It is referred to as " wiener " wave filter.
In superincumbent equation, h is the vector of filter coefficient, RxxIt is autocorrelation matrix (or blocks of data of reference frame), And RdxIt it is (between source frame and reference frame block data) cross-correlation matrix/row vector.
Although above-mentioned mathematical procedure can calculate filter coefficient, but by this method actual syndicated to Video coding Substantial amounts of problem is there is during in system.In the context H.264 and H.265 developed, it has been proposed that multiple suggestion Process based on picture (ALF), based on block (ALF), based on quaternary tree (QALF), Coefficient Space prediction, postfilter.
But, restriction be this process be computationally expensive (compared with separable filter, inseparable wave filter need More to calculate), be typically iteration, this need 3-12 complete iteration to calculate filter coefficient, and also usually Be required for each frame and send many coefficients (40-80 coefficient), this when differential coding possibly even for public middle layout Each frame of formula (" CIF ") resolution video spends the most about 300-600 bit, and sometimes provides limited gain (such as, owing to rounding off of coefficient is caused, typical coefficient is the matched well etc. of content).Additionally, certain methods sends grand Block or block based on quaternary tree map to indicate " unlatching " or " have closed " block of such wave filter;This process may be by expense Bit cost increases other 200-400 bit so that total bit cost of each frame is 500-1000 bit.This makes this System becomes present aspect to be expensive, inflexible and excessively complicated in overhead-bits.
Below with reference to document can find the extraneous information relevant with adaptive-filtering, every below with reference in document One is entirely incorporated into herein, for all purposes by reference:
T.Chujoh etc. " Block Based Adaptive Loop Filter ", ITU- Telecommunications Standardization Sector, VCEG, VCEG-AI18, Germany, July 2008.
" the Specification and experimental results of Quad-tree-such as T.Chujoh Based Adaptive Loop Filter ", ITU-Telecommunications Standardization Sector, VCEG, VCEG-AK22, Japan, April 2009.
Accompanying drawing explanation
Fig. 1 shows the exemplary video coder/decoder system according to an embodiment.
Fig. 2 shows some assemblies of the exemplary encoding device according to an embodiment.
Fig. 3 shows some assemblies of the exemplary decoding device according to an embodiment.
Fig. 4 shows according to an embodiment for generating " off-line " routine of code book, and described code book represents multiple infringement Compensate (" IC ") filter coefficient set.
Fig. 5 a-g shows the coefficient-pixel of the various wave filter used in compensation for loss and damage filters according to each embodiment Position.
Fig. 6 a-6d shows the exemplary system of each wave filter used in compensation for loss and damage filters according to each embodiment Number.
Fig. 7 shows the video encoder including IC code book wave filter encoder according to an embodiment.
Fig. 8 shows the Video Decoder including IC code book wave filter decoder according to an embodiment.
Fig. 9 shows the IC code book wave filter encoder according to an embodiment.
Figure 10 a-f shows the various compositions of the picture that can apply each embodiment.
Figure 11 shows the IC code book wave filter decoder according to an embodiment.
Figure 12 shows the quick IC code book wave filter encoder according to an embodiment.
Figure 13 shows the IC wave filter codebook search routine according to an embodiment, such as, can be performed by encoder.
Figure 14 shows the quick IC code book wave filter encoder with frequency spectrum more new encoder according to an embodiment.
Figure 15 shows the IC code book wave filter decoder according to an embodiment.
Figure 16 a-b shows that the exemplary transform coefficients that each embodiment is used scans and shielding.
Figure 17 shows the quick IC code book with frequency spectrum more new encoder and renewable code book according to an embodiment Wave filter encoder.
Figure 18 shows the IC code book wave filter decoder according to an embodiment.
Figure 19 shows the exemplary bit for IC code book wave filter encoder according to each embodiment Stream grammer.
Figure 20 a-b shows the sequence according to the picture in the video flowing of each embodiment or frame.
Figure 21 shows according to an embodiment for using code book that the IC wave filter of the picture of video is encoded into bit The routine of stream.
Figure 22 shows that the IC filter coefficient set according to an embodiment updates subroutine.
Figure 23 shows according to an embodiment for using one or more code book by the IC wave filter of video pictures It is encoded into bit stream the routine that it is updated.
Figure 24 shows according to an embodiment for using one or more code book to the video pictures from bit stream The routine that is decoded of IC wave filter.
Detailed description of the invention
Detailed description of the invention major part subsequently is the symbol of process and the operation carried out according to conventional computer components Representing and be given, conventional computer components includes processor, the memory storage device for processor, the display that connected Equipment and input equipment.Additionally, these processes and operation can utilize the conventionally calculation unit in heterogeneous distributed computing environment Part, including remote file server, computer server and memory storage device.In these conventional distributed computing components Each can be accessed via communication network by processor.
Repeatedly employ phrase " in one embodiment ", " in various embodiments ", " in certain embodiments " etc.. Such phrase is not necessarily referring to identical embodiment.Term " includes ", " having " and " comprising " is synonym, unless context is another Outer appointment.
General introduction
As discussed below, according to each embodiment, content-adaptive technology can be used to perform compensation for loss and damage filter Ripple device coefficient calculations and expression based on code book, in order to perform filtering, thus improve the coding effect in motion compensated video coding Rate.There is described herein for video being carried out compensation for loss and damage filtering to improve each of predictive efficiency of inter-frame video coding and be System and method.
Various technology described below can be efficient in terms of bit, this is because phase own with transmitting filter Ratio, can only send the index for storing wave filter in the codebook.With actual coefficient is encoded or differentially The filter coefficient sending coding is compared, and it is more efficient for encoding index.Although the quantity of the actual index sent takes Certainly in the quantity of subregion (partitions), but compared with the method before using similar divisional type, skill described below Art can be more efficient in terms of bit rate.
Technology described below may be provided for the computation complexity reduced, and this is not based on the scheme of code book not with existing With, it can avoid following demand: calculate different size (such as 5 × 5,7 × 7,9 × 9) simultaneously and/or type is (the most non-right Claim and point symmetry) wiener wave filter, and calculate iteratively will wherein application or do not apply the block of wave filter to map, and The most again calculate the wave filter of improvement of these sizes and type to select optimal candidate based on exhaustive rate-distortion optimisation. On the contrary, technology described below depends on process based on content and determines wave filter size and filter type, keeps simultaneously Use the ability (if desired) that block maps.Meanwhile, technology described below can search for code book efficiently to determine Good candidate.
Additionally, technology described below can provide motility, it is allowed to by using code book according to what it is today, provide simultaneously Ability (this and the transmission coefficient itself of the optimum filter that correction of more newly arriving with relatively little of bit frequency spectrum can obtain from code book Compare, remain more efficient).The precision that frequency spectrum updates can be controlled.It addition, technology described below allows more escape This some, or allow the wave filter of the correction being more likely to be appropriate for video sequence to be coded of by increase to expand Show code originally, thus improves the overall efficiency of codebook design.
Technology described below can also map at picture, sheet group (slice group), block, merge fast map or from Adapt to the upper work of subregion (such as binary tree, quaternary tree or content isolated area).With only drive by rate-distortion optimisation its Its technology is compared, and the ability that technology described below works together with content independent partitions also reduces complexity, and improves Precision.
The details of wave filter codebook design and the effective search of code book are described below.Use described codebook design side Method, and for the codebook size recommended, technology described below can be that large-scale video content brings good infringement Compensation filter, improves predictive efficiency.
In various embodiments, technology described below can be incorporated in general Video coding, and its use is traditional The picture structure being made up of I-, P-, B-coding of graphics.In other embodiments, technology described below can be incorporated into use The video of other structure (such as layering B-picture, unidirectional B-picture and/or B-picture are replaced) in addition to I-and P-picture In coding.
In various embodiments, compensation for loss and damage based on code book (" IC ") described below filtering can serve as not showing institute In the loop of the picture of filtering, the loop of the picture that (in-loop), display are filtered is interior and/or only shows the picture filtered Only post processing.In addition to visual quality is improved, its use in loop can also allow for compression efficiency gain.
The description of the embodiment with reference to shown in accompanying drawing the most in detail.Although describing in conjunction with accompanying drawing and associated description Embodiment, but it is not intended to scope is limited to embodiment disclosed herein.On the contrary, it is intended to be to contain all replacements, amendment And the equivalent form of value.In alternative embodiments, the combination of extra equipment or shown equipment can be increased or combine, and also Scope is not limited to presently disclosed embodiment.
Fig. 1 shows the exemplary video coder/decoder system 100 according to an embodiment, including uncoded video Source 115, encoding device 200 (seeing Fig. 2, be discussed below), the decoding device 300 with optional external display 125 (see figure 3, be discussed below) and the source 120 of optional encoded video.In certain embodiments, encoding device 200 and/or decoding set One or two in standby 300 may be coupled to network 150.In certain embodiments, decoding device 300 and encoding device 200 Individual equipment can be included.
In certain embodiments, it is also possible to there is other server and/or equipment (not shown).Such as, implement at some In example, Distributor (not shown) can distribute encoded video via network 150 to decoding device 300.
In certain embodiments, encoding device 200 can be via network 150, storage area network (" SAN "), at a high speed string Row bus and/or the source 115 via other suitable communication technology with uncoded video communicate.Similarly, real at some Executing in example, decoding device 300 can be via network 150, storage area network (" SAN "), high-speed serial bus and/or via it Its suitable communication technology communicates with the source 120 of encoded video.In certain embodiments, encoding device 200, decoding set The source 115 of standby 300, the source 120 of encoded video and/or uncoded video can include one or more repetition and/or point The physically or logically equipment of cloth.
In various embodiments, network 150 can include the Internet, LAN (" LAN "), wide area network (" WAN "), honeycomb Data network and/or other data network.In many examples, compared with shown, more coding can be there is and set For 200, decoding device 300, the source 120 of encoded video and/or the source 115 of uncoded video.
Fig. 2 shows some assemblies of the exemplary encoding device 200 according to an embodiment.In certain embodiments, Compared with those assemblies shown in Fig. 2, encoding device 200 can include more assembly.But, in order to open the most real Execute example, it is not necessary that illustrate in these general general components is whole.As shown in Figure 2, encoding device 200 includes optional Network interface 230, be used for being alternatively coupled to network 150.
Encoding device 200 also include processing unit 210, memorizer 250, optional outut device and/or display 240 with And optional input equipment 270 (such as, keyboard, mouse, remote controller, stick etc.), all together with optional network interface 230 It is interconnected at together via bus 220.Memorizer 250 generally comprises random access memory (" RAM "), read only memory (" ROM ") and permanent mass storage device, such as disc driver.Memorizer 250 stores IC wave filter code book 265, is used for Have the video encoder of IC filtering program code 260 and IC wave filter code book generating routine 400 (see Fig. 4, below Discuss), IC wave filter codebook search routine 1300 (seeing Figure 13, be discussed below), IC wave filter code book coding routine (see figure 21, be discussed below) and many code books IC wave filter code book coding routine (seeing Figure 23, be discussed below).Additionally, memorizer 250 is also Storage operating system 255.Can use and non-transitory computer-readable storage media 295 (such as, floppy disk, band, DVD/CD- ROM drive, storage card etc.) the driving mechanism (not shown) that is associated from computer-readable recording medium 295 by these software Assembly is loaded in the memorizer 250 of encoding device 200.In certain embodiments, it is also possible to via optional network interface 230 Rather than load component software via computer-readable recording medium 295.
In certain embodiments, encoding device 200 could be included for what the source 115 with uncoded video communicated Special purpose interface 235, such as high-speed serial bus etc..In certain embodiments, encoding device 200 can be via network interface 230 Communicate with the source 115 of uncoded video.In other embodiments, the source 115 of uncoded video may be located at memorizer 250 In.
Although it have been described that generally conform to the exemplary encoding device 200 of conventional universal computing device, but coding Equipment 200 can be any one in the greater number of equipment that can encode video, such as video recording apparatus, Video coprocessor and/or accelerator, personal computer, game console, Set Top Box, handheld computer, cell phone or Other suitable equipment any.
Fig. 3 shows some assemblies of the exemplary decoding device 300 according to an embodiment.In certain embodiments, Compared with those assemblies shown in Fig. 3, decoding device 300 can include many more assemblies.But, illustrate in order to open Property embodiment, it is not necessary to illustrate in these general general components is whole.As shown in Figure 3, include can for decoding device 300 The network interface 330 of choosing, is used for being alternatively coupled to network 150.
Decoding device 300 also include processing unit 310, memorizer 350, optional outut device and/or display 340 with And optional input equipment 370 (such as, keyboard, mouse, remote controller, stick etc.), all together with optional network interface 330 It is interconnected at together via bus 320.Memorizer 350 generally comprises random access memory (" RAM "), read only memory (" ROM ") and permanent mass storage device, such as disc driver.Memorizer 350 stores the video for having IC filtering The program code 360 of decoder, IC wave filter code book 365 (duplicate copies of IC wave filter code book 265), for IC filter code book Program code 360 and IC wave filter code book decoding routine (seeing Figure 24, be discussed below) of decoder is additionally, memorizer 350 Also storage operating system 355.Can use and non-transitory computer-readable storage media 395 (such as, floppy disk, band, DVD/ CD-ROM drive, storage card etc.) the driving mechanism (not shown) that is associated is soft by these from computer-readable recording medium 395 Part assembly is loaded in the memorizer 350 of decoding device 300.In certain embodiments, it is also possible to via optional network interface 330 rather than load component software via computer-readable recording medium 395.
In certain embodiments, decoding device 300 could be included for what the source 120 with encoded video communicated Special purpose interface 335, such as high-speed serial bus etc..In certain embodiments, decoding device 300 can be via network interface 330 Communicate with the source 120 of encoded video.In other embodiments, the source 115 of encoded video may be located at memorizer 350 In.
Although it have been described that generally conform to the exemplary decoding device 300 of conventional universal computing device, but decoding Equipment 300 can be any one in the greater number of equipment can being decoded video, such as video playback apparatus, Personal computer, game console, Set Top Box, handheld computer, cell phone or other suitable equipment any.
Fig. 4 shows according to an embodiment for generating " off-line " routine 400 of code book, and described code book represents multiple IC Filter coefficient set.In certain embodiments, routine 400 may spend some hours or perform over the most some skies.
In square frame 405, it is thus achieved that include the test set of multiple (such as 25 to 50) video sequence, each video sequence There is the frame of some (such as 20-100).In certain embodiments, some or all in these video sequences are permissible Select from one or more video sequences being pre-existing in.In other embodiments, these video sequences can be developed In some or all to be used together with routine 400.In certain embodiments, some or all in these video sequences can To have various characteristic, such as high-contrast, low contrast, high motion, harmonic motion, bright, dark etc..In square frame 410, select one Individual or multiple wave filter sizes.Such as, in one embodiment, can select in the wave filter size of 5 × 5,7 × 7 and 9 × 9 One or more.In square frame 415, select one or more target bit rate.Such as, in one embodiment, Ke Yixuan That selects in low, medium and high bit rate is one or more.
The square frame 420 started the cycle over starts, in turn accesses (address) each wave filter size.Follow starting The square frame 425 of ring starts, in turn processes each video sequence.The square frame 430 started the cycle over starts, in turn accesses Each selected bits rate.The square frame 435 started the cycle over starts, in turn processes each picture in current video sequence Or frame.
In square frame 440, photo current is encoded by video encoder.In square frame 445, for photo current, filter Ripple device size and bit rate, filter by coded picture and original uncoded video are compared to calculating Wiener Device.In square frame 450, at least temporarily with the Wiener wave filter calculated is stored into candidate's Wiener wave filter.
In the square frame 455 of end loop, routine 400 loops back square frame 435 to process next in current video sequence Picture (if present).In the square frame 460 of end loop, routine 400 is circulated back to square frame 430 to process following bit Rate (if present).In the square frame 465 of end loop, routine 400 loops back square frame 425 to process next video sequence (if present).
Once have been directed towards all pictures that each bit rate has processed in all video sequences, in square frame 470, to coming The subset (such as 256) of the different Wiener wave filter in the middle of the candidate's Wiener wave filter calculated selects, with Cover a series of filter characteristics (such as from smooth to sharp-pointed) from low contrast to high-contrast.
In square frame 475, such as according to the son of the sequential storage institute selecting filter from smooth to sharp-pointed (seeing Fig. 6 a-d) Collection (such as, according to center coefficient (or other exemplary parameter) with ascending order order), and this subset is stored in for current filter In the code book of the filter coefficient of ripple device size.
In the square frame 480 of end loop, routine 400 be circulated back to square frame 420 with process next wave filter size (if If existence).After having generated the code book for each wave filter size, routine 400 terminates in square frame 499, institute The code book generated is available for various encoder and uses.
Fig. 5 a-d respectively illustrates the IC filter coefficient set of the inseparable wave filter of 3 × 3,5 × 5,7 × 7 and 9 × 9 Close, may be used for compensation for loss and damage filtering the most in various embodiments.Such as, as illustrated in fig. 5 a, in the feelings of 3 × 3 filtering In condition, create filtered decoding picture pixel by pixel by the filter coefficient 505 of application 3 × 3, wherein, with in feel concerned about Number c11The pixel of alignment is that the weighted value using this pixel and 5 pixels of surrounding is filtered, and wherein weighted value is by wave filter Coefficient is given.Owing to wave filter is inseparable, so 9 coefficients can have unique value.Similarly, as illustrated in figure 5d , when the asymmetric filtering of 9 × 9 for, pixel to be filtered and center coefficient C44Alignment, and except this picture Beyond element, also using 50 pixels of surrounding, the most each pixel of being filtered is the weighted array of 51 pixels;Filtered Pixel is written out the picture as the decoding individually filtered, ginseng that is that it can serve as prediction and that may even show Examine.In various embodiments, can the coefficient 510 and 515 shown in application drawing 5b and Fig. 5 c in a similar fashion.
Equally, Fig. 5 e-h respectively illustrates the coefficient of the point symmetry wave filter of 3 × 3,5 × 5,7 × 7 and 9 × 9, such as, exist Some embodiments may be used for compensation for loss and damage filtering.Such as, as shown in fig. 5e, in the case of 3 × 3 filtering, pass through Application 3 × 3 filter coefficients 525 create the picture of filtered decoding pixel by pixel, wherein, with center coefficient c11Alignment Pixel is that the weighted value using this pixel and 5 pixels of surrounding is filtered, and wherein weighted value is given by filter coefficient.By It is point-symmetric in wave filter, so only 5 coefficient c00、c01、c02、c10、c11It is likely to be of unique value, and other coefficient is simple Diagonal angle, single ground is symmetrical.Similarly, as shown in Fig. 5 h, for the point symmetry for 9 × 9 filters, only 41 coefficients are only One, other is that diagonal angle is symmetrical.Pixel to be filtered and center coefficient c44Alignment, and also use in addition to this pixel 50 pixels around, each filtered pixel is the weighted array of 51 pixels;The pixel filtered is written out as list The picture of only decoding filtered, reference that is that it can serve as prediction and that may even show.In each embodiment In, can the coefficient 530 and 535 shown in application drawing 5f and Fig. 5 g in a similar fashion.
Fig. 6 a shows the exemplary system of inseparable " sharply " wave filter of use in compensation for loss and damage filters 9 × 9 Number.Fig. 6 b shows the exemplary coefficient of point symmetry " sharply " wave filter of use in compensation for loss and damage filters 9 × 9.Fig. 6 c Show the exemplary parameter of inseparable " smooth " wave filter of use in compensation for loss and damage filters 9 × 9.Fig. 6 d shows The exemplary coefficient of point symmetry " smooth " wave filter of 9 × 9 used in compensation for loss and damage filters.
Fig. 7 shows and have employed compensation for loss and damage in motion compensated prediction and loop based on code book according to an embodiment The general block diagram of the interframe coder 700 of filtering.There is provided one or more to block index 701 and scene content analysis device 703 Frame of video (vidfrms).Scene content analysis device 703 is analyzed one or more frame of video (vidfrms) and exports instruction scene The mark (scnchg) changed and the signal of instruction picture characteristics, the such as regional edge in scene contrast (contr) and frame Boundary (regbnd).One or more frame of video (vidfrms) are divided in bulk (blks), these blocks one time by block index 701 It is imported into difference engine 733 individually, and subtracts each other with corresponding prediction signal block (pred).Block (blks) is also provided to motion Estimator 728 (being discussed below).After carrying out difference at difference engine 733, changer 235 is by produced residual error (res) just Being converted into frequency domain representation, thus produce coefficient block (tcof), coefficient block (tcof) is subsequently sent to quantizer 738, thus produces Raw quantization parameter block (qcf), quantization parameter block (qcf) is sent to entropy coder 718 and inverse quantizer 740.Inverse quantizer 740 The coefficient block of de-quantization (tcof ') is sent to inverse converter 743.At adder 745, from the prediction of motion compensation loop Block (pred) is added on the residual block (res ') of the de-quantization of the output of inverse converter 743.Adder 745 exports this locality The block (rec) of decoding, the block (rec) of local decoding is subsequently sent to frame assembler and deblocking filter processor 748, frame group Dress device and deblocking filter processor 748 reduce block effect.The output of frame assembler and the filter processor 748 that deblocks Block (recd) forms the frame being processed into compensation for loss and damage reconstructed frame (recf) by IC code book wave filter encoder 750, compensation for loss and damage weight Structure frame (recf) is used as exercise estimator 728 and the reference frame of motion compensated predictor 730.Exercise estimator 728 calculates original The block (blks) of frame of video is relative to the motion vector (mv) between compensation for loss and damage reconstructed frame (recf).In certain embodiments, Motion vector calculation process includes calculating the integer picture point vector for each piece, is that sub-picture point vector (such as, arrives 1/ afterwards 4 picture point resolution).The entropy coder 718 conversion coefficient (qcf), the motion vector (dmv) of difference and other data to quantifying (including the one or more signals 753 from IC code book wave filter encoder 750) encodes, thus generates the bit of compression Stream 728.In certain embodiments, entropy coder 718 can also provide bit count to IC code book wave filter encoder 750 (btcnt)。
Fig. 8 shows corresponding decoder, and works similarly with the local decoding loop 755 at encoder 700.Tool Body ground, the bit stream 801 of the compression that will be decoded is input to entropy decoder 803, and entropy decoder 803 is to being supplied to IC code The coefficient block (qcf) of the quantization of this wave filter decoder 820 (see, e.g. Figure 10,13,15,18, be discussed below), difference are transported Moving vector (dmv) and other data (including control signal 818) are decoded.Inverse quantizer 813 is to quantization parameter block subsequently (qcf) carry out re-quantization, produce the coefficient (tcof ') of de-quantization.Inverse converter 815 is by the coefficient of de-quantization (tcof ') subsequently Inversion swaps out frequency domain, produces the residual block (res ') of decoding.Adder 823 is plus by using corresponding sports vector (mv) to obtain Motion-compensated prediction block (pred).The video (rec) of produced decoding at frame assembler and deblocks in filter processor 825 By de-blocking filter.Just as at encoder 700, the block of the output of frame assembler and the filter processor 825 that deblocks (recd) define and be processed into the frame of compensation for loss and damage reconstructed frame (recf) by IC code book wave filter decoder 820, compensation for loss and damage weight Structure frame (recf) is used as the reference frame of motion compensated predictor 830.The working method of motion compensated predictor 830 is similar to coding Motion compensated predictor 730 at device 700, such as it uses similar sub-picture point interpolation device.
Fig. 9 shows the IC code book wave filter encoder 900 according to an embodiment.Frame of video (not shown) is divided into Block (blks), these blocks are imported into encoder 900 one at a time.Be supplied to encoder 900 also has reconstruction signal (recd)。
Space dispenser and zonal coding device 905 determine that the subregion of candidate maps (cand.ptmap) and type (cand.pttyp), they are provided to bit rate/distortion compromise analyzer 908.Such as, in one embodiment, candidate's Subregion maps (cand.ptmap) can include that binary maps (if only two kinds regions or block type exist), ternary maps (existing in the case of three kinds of regions or block), two binary mappings etc..
Bit rate/distortion compromise analyzer 908 selects subregion to map (selected ptmap or simply ptmap) and subregion Type (selected pttyp or simply pttyp), they are returned to space dispenser and zonal coding device 905, and are passed to IC filter processor 915.Additionally, in certain embodiments, subregion mapping (ptmap) can be coded by entropy device 918 and be encoded into ratio Special stream 928, to be sent to decoder.But, with its primitive form (such as, 0/1 shields or even more than 0/1 shielding), point District maps (ptmap) and is transmitted in original form being likely to be of relatively high bit cost.Thus, in many examples, Entropy coder 918 generates subregion and maps the compressed version (be referred to as encoded subregion and map (ptmapc)) of (ptmap) and by it It is encoded in bit stream 928.If use lossless compress, then decoder can map (ptmapc) to encoded subregion to be carried out Decoding, and derive definite subregion mapping (ptmap) to use at decoder.In various embodiments, for subregion Map (ptmap) carry out encoding generating encoded subregion map the compression method of (ptmapc) can include Huffman coding, The distance of swimming/length coding, arithmetic coding or other similar lossless coding technique.Additionally, in some cases, lossless compress may It is not enough to reduce the bit cost sending subregion mapping (ptmap).In such a case, can be raw via lossy compression Encoded subregion is become to map (ptmapc).
According to each embodiment, space dispenser and zonal coding device 905 and/or bit rate/distortion compromise analyzer 908 Various different Fractionation regimen can be used.In various embodiments, can according to frame, according to region or fragment, according to sheet group, According to single sheet, according to each quaternary tree of frame or binary tree subregion, according to the macro block of frame or block group, according to tile, press According to the segmentation of tile and/or (it can allow to work as from multiple available filters by using tile, segmentation, macro block or block to map In select) application method described herein.
Such as, Figure 10 a-f shows according to each embodiment for dividing the various Fractionation regimen of picture.Such as, Figure 10 a Showing segmentation based on picture, wherein, wave filter coding only switches over (switch) on the basis of by picture.Figure 10 b Show based on sheet or the segmentation of sheet group so that at sheet group " A " upper application the first wave filter collection, and in the upper application of sheet group " B " the Two wave filter collection.Figure 10 c shows that block maps, and which wave filter it identifies is used for for block type " A " and which wave filter Block type " B ".Figure 10 d shows and picture is divided into tile, and tile is further divided into binary tree subregion so that each Based on rate-distortion optimization, (" the binary tree editing of RDO ") is marked as " A " or " B ", indicates them can correspondingly use first Or the second wave filter collection.Equally, Figure 10 e shows and picture segmentation becomes tile, and is divided into by tile labeled further Two kinds of quaternary tree subregions for " A " or " B " so that they can correspondingly use the first or second wave filter collection.Finally, figure 10f shows by merging sheet or merging block or merge binary tree or merge quaternary tree subregion or pass through purely The region that space divides and obtains.
Refer again to IC code book wave filter encoder 900 and Fig. 9, block (blks) be provided to picture postpone 925 and with After be supplied to cross correlation vector computer 923, cross correlation vector computer 923 calculates source frame blocks of data (blks) and reconstructed frame block Cross-correlation matrix between data (recd)/row vector (Rdx).Autocorrelation matrix computer 920 calculates reconstructed frame blocks of data (recd) autocorrelation matrix (Rxx)。
IC wave filter codebook search processor 910 is according to picture characteristics (the such as scene changes (scnchg) of input picture With contrast (contr)) determine that wave filter size (fsz) (such as, 5 × 5,7 × 7,9 × 9 etc.) and pattern (md) are (such as, non- Symmetrical or point-symmetric).(seeing Fig. 5 a-h, be discussed below) such as, when input picture be " low contrast " time, it may be determined that Big wave filter size (such as 9 × 9).On the other hand, when inputting picture and being " high-contrast ", if coding quality Gao Zeke To determine little wave filter size (such as 5 × 5), if or coding quality be in wait until the lowest, may determine that medium filtering Device size (such as 7 × 7).For medium-contrast scene, may determine that medium wave filter size when coding quality is enough (such as 7 × 7), or big wave filter size (such as 9 × 9) is may determine that when coding quality is low.In certain embodiments, Transmitting filter size information on frame-basis.
Use cross-correlation matrix/row vector (Rdx) and autocorrelation matrix (Rxx), IC wave filter code book look-up table (" LUT ") Scan in IC wave filter code book with address logic 913, with location at least substantially with the target filter calculated IC filter coefficient set (icfc) that coefficient sets meets.
In certain embodiments, codebook search relates to each single filter coefficient set is applied to whole weight Structure frame blocks of data (recd).The filter coefficient set won is selected as producing minimum relative to original source frame blocks of data (blks) the filter coefficient set of absolute difference sum (" SAD ").
In other embodiments, codebook search relates to calculating in comparison measuring or matching criterior, such as wave filter code book The mean square error (" MSE ") of IC wave filter, and select the wave filter that comparison measuring (such as, MSE, SAD etc.) is minimized Coefficient sets.Such as, (that is, primitive frame pixel d (n) and filtered reconstructed frame pixel y (n) are (by reconstruct for two frame of video Frame pixel x (n) is filtered obtaining)) between the calculating of MSE relate to: that tries to achieve between the two frame of video is the poorest Quadratic sum, and divided by the pixel quantity in frame of video, as follows:
SSE=Σ (d (n))2+Σ(y(n))2-2Σd(n)y(n)
Using wave filter h () to be filtered signal x () thus produce the general operation of filtered output y () can be by It is written as:
Y (n)=Σ h (k) x (n-k)
Replace the y () in mean square error sum (" SSE ") formula,
SSE=Σ (d (n))2+ΣΣ(h(k)x(n-k))2-2Σd(n)(∑h(k)x(n-k))
The formula that this SSE formula is actually discussed with the wave filter computing interval is identical, i.e.
In the case of given one group of wave filter, in order to try to achieve the SSE of minimum, the wave filter that location generation is following:
Additionally, autocorrelation matrix RxxOr cross-correlation matrix RdxIt is not rely on wave filter h, and it practice, RxxAnd RdxRight All calculate in advance (during calculating optimal IC wave filter) for given frame, and thus can reuse.
To compare with direct SSE complexity of the calculation (in multiplication operating aspect) based on relevant SSE computational methods Relatively, the magnitude of the complexity of actual filtering is that hsz × vsz × taps (such as, is for CIF video and 81 tap filtering 352 × 288 × 81=8,211,456), and the magnitude of the complexity of correlation technique be taps × taps+taps (such as, for For CIF video and 81 tap filtering be 81 × 81+81=6642 or than few more than 1000 times of actual filtering).
Refer again to Fig. 9, IC wave filter code book LUT and address logic 913 thus scan in IC wave filter code book To position IC filter coefficient set (icfc) making MSE (or other matching criterior) minimize.The IC wave filter positioned The index (icfci) of collection (icfc) and determined by wave filter size (fsz) and pattern (md) be provided to entropy coder 918, to be included in bit stream 928.Using IC filter coefficient set (icfc) positioned, IC filter processor 915 is right Reconstruction signal (recd) mapped (ptmap) and being filtered at least partially indicated by divisional type (pttyp) by subregion To obtain the reconstruction signal (recf) of compensation for loss and damage, it is the most loyal compared with the reconstruction signal (recd) that non-damaging compensates In original source signal (being represented by source frame blocks of data (blks)).
Figure 11 shows the IC code book wave filter decoder 1100 corresponding with encoder 900 discussed above.To be solved The compression bit stream 1101 of code is imported into entropy decoder 1103, and encoded subregion is mapped by entropy decoder 1103 (ptmapc), the IC filter coefficient set in divisional type (pttyp), wave filter size (fsz) and code book that size is fsz Index (icfci) be decoded.
Encoded subregion maps (ptmapc) and divisional type (pttyp) is provided to space partition zone decoder 1105, Space partition zone decoder 1105 maps (ptmapc) and is decoded and the subregion of decoding is mapped (ptmap) encoded subregion It is supplied to IC filter processor 1108.
Wave filter size (fsz) and index (icfci) are provided to IC wave filter code book LUT and address logic 1113, its Obtain IC filter coefficient (icfc) from code book, and IC filter coefficient (icfc) is supplied to IC filter processor 1108.IC IC filter coefficient (icfc) is applied to reconstruction signal (recd) by filter processor 1108, thus generates the reconstruct of compensation for loss and damage Signal (recf).
Figure 12 shows the quick IC code book wave filter encoder 1200 according to an embodiment.IC code book wave filter encodes Device 1200 is similar to IC code book wave filter encoder 900 discussed above in many aspects.But, encode with IC code book wave filter Unlike device 900, Wiener filter calculator 1230 uses cross-correlation matrix/row vector (Rdx) and autocorrelation matrix (Rxx) calculate the target filter coefficient sets of the Wiener wave filter with given wave filter size (fsz) (calc.icfc).Approximate match computer 1233 uses target Wiener filter coefficient (calc.icfc) to come at IC wave filter Code book positions and the IC filter coefficient set of target filter coefficient sets (calc.icfc) approximate match calculated Close.Such as, in one embodiment, the IC filter coefficient set of approximate match is confirmed as having and target filter coefficient The immediate coefficient sets representing coefficient (such as center coefficient) of set (calc.icfc).In one embodiment, search Immediate center coefficient performs via binary search algorithm, this is because the coefficient sets in IC wave filter code book is According to center coefficient sequence.
Once approximate match computer 1233 located the IC filter coefficient set of approximate match, IC wave filter code This search processor 1210 has determined that neighbours radius R (such as, 8 or 16), and neighbours radius R has 2R+1 filtering for selection The comparable group of device coefficient sets, before this comparable group includes the IC filter coefficient set of the approximate match in IC wave filter code book R wave filter and R wave filter afterwards.
IC wave filter codebook search processor 1210 search for subsequently neighbours' comparable group with determine so that MSE (SAD or other Matching criterior) minimum IC filter coefficient set (icfc).In certain embodiments, IC wave filter codebook search processor The possible modification of 1210 also search for each filter coefficient set in neighbours' comparable group up to eight (rotates and inverts or mirror Picture).
Once located good IC filter coefficient set (icfc), just provide index to entropy coder 1218 (icfci) (wave filter size (fsz) determined by and and pattern (md)), with (discussed below together with shown in Figure 19 One or more extra bit stream descriptors) it is included in bit stream 1228.
Figure 13 shows the IC wave filter codebook search routine 1300 according to an embodiment, such as, can be held by encoder OK.In square frame 1305, routine 1300 obtains uncoded picture or the frame of video.In square frame 1310, routine 1300 uses and regards Frequently picture is encoded by encoder with certain bit rate.In certain embodiments, video encoder can be interframe coder. In other embodiments, video encoder can be intra encoder.In square frame 1315, routine 1300 can be via in loop Local decoder the coded picture at video encoder is carried out local decoding.
In square frame 1320, routine 1300 determines the contrast (such as, low, medium and high) of original uncoded picture. In square frame 1325, routine 1300 according to determined by contrast select wave filter size (fsz).Such as, an enforcement In example, routine 1300 can select little wave filter size (such as, 5 × 5) when contrast height, in selecting when contrast is low Deng or big wave filter size (such as, 7 × 7 or 9 × 9), and otherwise select big wave filter size (such as 9 × 9).
In square frame 1330, routine 1300 according to coding bit rate, determined by contrast or other similar factor Some or all select filter type (such as, asymmetrical or symmetrical).
In square frame 1335, routine 1300 calculates the selected size (fsz) of the picture of this locality decoding relative to original image The coefficient of target Wiener wave filter, wherein, the picture decoded this locality is used as reference signal, and by original uncoded figure Sheet is used as desired signal.In square frame 1340, routine 1300 selects to include one or more candidate's IC wave filter from code book Comparable group.In certain embodiments, can be by (such as, filtering with target Wiener from the wave filter of code book location approximate match Ripple utensil has the IC wave filter of close or immediate center coefficient), and select the wave filter of approximate match any subsequently Some (such as 8 or 16) adjacent filter on side, selects one or more candidate's IC wave filter.Implement at some In example, neighbours' comparable group can also include one or more modification or the conversion of each candidate's IC wave filter.Such as, asymmetric Wave filter there are four modification rotated and four mirrored variant, and the wave filter of symmetry has four rotated version.
Starting in the square frame 1345 started the cycle over, it is each that routine 1300 processes in one or more candidate's IC wave filter Individual, square frame 1350 determines the error estimation (such as, SAD, MSE etc.) of current candidate IC wave filter, and from square frame 1355 Loop back to process next candidate's IC wave filter (if present).In square frame 1360, routine 1300 selection has low Determined by candidate's IC wave filter of error estimation.In certain embodiments, routine 1300 can select have minimum institute really Candidate's IC wave filter of fixed error estimation.
In decision block 1365, routine 1300 determines whether IC wave filter selected compared with not having IC wave filter represents Improve.Such as, in one embodiment, routine 1300 determines and carries out the picture of decoding after IC filtering with selected IC wave filter Whether the error estimation (such as SAD, MSE etc.) of (compared with original uncoded picture) is less than the solution not carrying out IC filtering The error estimation of code picture.If selected IC filtering does not provide improvement, then in square frame 1370, in encoded bit stream The wave filter of one bit sends labelling (fsnt) and is configured to 0.On the other hand, if selected IC wave filter provides improvement, that In square frame 1375, the wave filter of the bit in encoded bit stream sends labelling (fsnt) and is configured to 1, and in side In frame 1380, encoded bit stream arranges the index (icfci) of the selected IC wave filter from code book.In some embodiments In, routine 1300 can also arrange modification or conversion (such as, the rotation of the selected IC wave filter of instruction of the selected IC wave filter of instruction Or mirrored variant) signal (flt_ornt).(seeing Figure 19 discussed below to obtain being explained in more detail of bitstream syntax) Routine 1300 terminates in square frame 1399.
Figure 14 shows the quick IC code book wave filter encoder with frequency spectrum more new encoder according to an embodiment 1400.IC code book wave filter encoder 1400 is similar to quick IC code book wave filter encoder discussed above in many aspects 1200.But, unlike IC code book wave filter encoder 1200, IC can be filtered by IC code book wave filter encoder 1400 The renewal of the IC filter coefficient (icfc) selected by device code book LUT and address logic 1413 carries out spectrum coding, as follows.
Wiener filter calculator 1430 uses cross-correlation matrix/row vector (Rdx) and autocorrelation matrix (Rxx) count Calculator has the target filter coefficient sets of the Wiener wave filter of given wave filter size (fsz) and pattern (md) (calc.icfc).As in IC code book wave filter encoder 1200, approximate match computer 1433 uses target Wiener to filter Device coefficient (calc.icfc) position in IC wave filter code book approx with the target filter coefficient sets calculated (calc.icfc) the IC filter coefficient set mated.
But, unlike IC code book wave filter encoder 1200, Wiener filter calculator 1430 is also by target Wiener filter coefficient (calc.icfc) is supplied to difference engine 1435, in difference engine 1435, and these coefficients and corresponding institute IC filter coefficient (icfc) is selected to carry out difference.
After carrying out difference at difference engine 1435, produced coefficient difference (icfcd) adjusted value is transformed device 1438 Direct transform becomes frequency domain representation, thus produces the block of the difference coefficient of conversion, and it is subsequently sent to quantizer 1440.Quantizer The block of the difference coefficient of conversion is quantized into the difference of the conversion of quantization according to quantization parameter (qp is sometimes referred to as quantization scale) by 1440 Dividing the block of coefficient, it is sent to scanning device 1443 and inverse quantizer 1445.Scanning device 1443 according to inverse scan order to quantization The block of difference coefficient of conversion be scanned (seeing Figure 16 a, be discussed below), and by the difference system of the transition coding after scanning Number (tcod.iccof) is supplied to entropy coder 1418, to be encoded in bit stream 1428 (refer to if being updated signal (updt) If showing) (being discussed below).
Inverse quantizer 1445 carries out de-quantization according to the block of difference coefficient of the quantization parameter (qp) conversion to quantifying, and will The difference coefficient of the conversion of de-quantization is sent to inverse converter 1448.At adder 1450, the difference of the de-quantization of inverse transformation Coefficient is added on the selected IC filter coefficient (icfc) of IC wave filter code book LUT and address logic 1413, to be formed The IC filter coefficient (icfcu) updated.
Controller 1455 sends more new signal (updt) to switch 1453, and this direction switch signal 1453 should filter to IC Processor 1415 sends selected IC filter coefficient (icfc) or the IC filter coefficient (icfcu) updated.In order to determine More new signal (updt), controller 1455 uses the signal (pictyp) (such as, I-, P-or B-picture/mb-type) of picture/mb-type With importance signal (sigf).Importance signal (sigf) is the threshold value choosing carried out at IC code book wave filter encoder 1400 Selecting, whether the selected difference between IC filter coefficient (icfc) and target IC filter coefficient (calc.icfc) of its instruction exceedes Predetermined importance threshold value.If difference has exceeded threshold value, then more new signal (updt) is to switch 1453 instruction IC Filtering Processing Device 1415 should use the IC filter coefficient (icfcu) of renewal.Additionally, more new signal (updt) is also compiled by entropy coder 1418 Code is in bit stream 1428.
Figure 15 shows the IC code book wave filter decoding corresponding with IC code book wave filter encoder 1400 discussed above Device 1500.Compression bit stream 1501 is imported into entropy decoder 1503, and encoded subregion is mapped by entropy decoder 1503 (ptmapc), the difference of the transition coding after divisional type (pttyp), wave filter size (fsz), more new signal (updt), scanning The index (icfci) of the IC filter coefficient set in point coefficient (tcod.iccof) and code book that size is fsz is decoded.
Encoded subregion maps (ptmapc) and divisional type (pttyp) is provided to space partition zone decoder 1505, empty Between subregion decoder 1505 to encoded subregion map (ptmapc) be decoded, and by decoding subregion map (ptmap) carry Supply IC filter processor 1508.
Wave filter size (fsz) and index (icfci) are provided to IC wave filter code book LUT and address logic 1510, IC Wave filter code book LUT and address logic 1510 obtain IC filter coefficient (icfc) from code book.
Inverse scan device 1513 according to inverse scan order to scanned differential transformation coding coefficient (tcod.iccof) enter Row inverse scan (sees Figure 16 a, be discussed below), to obtain the block of the difference coefficient of the conversion quantified.Inverse quantizer 1515 basis The block of the difference coefficient of the quantization parameter (qp) conversion to quantifying carries out de-quantization, and is sent out by the difference coefficient of the conversion of de-quantization Deliver to inverse converter 1520.At adder 1523, the difference coefficient adjusted value of the de-quantization of inverse transformation is added to selected IC filter Ripple device coefficient (icfc) is to form the IC filter coefficient (icfcu) updated.
Switch 1525 (according to more new signal (updt)) determine provides selected IC wave filter system to IC filter processor 1508 Number (icfc) or the IC filter coefficient (icfcu) updated.The coefficient provided is applied to weight by IC filter processor 1508 Structure signal (recd), thus produce the reconstruction signal (recf) of compensation for loss and damage.
Figure 16 a shows the example that IC code book wave filter decoder 1500 and IC code book wave filter encoder 1400 is used Property conversion coefficient inverse scan.In general, usually change is scanned starting from the order of the z font of DC coefficient (such as, 1605) Change coefficient.But, in shown inverse scan, to end at the sequential scan conversion coefficient of the z font of DC coefficient 1605.
Figure 16 b shows and can be made in conjunction with IC code book wave filter decoder 1800 and IC code book wave filter encoder 1700 Coefficient inverse scan and use exemplary transformations shielding.As shown in Figure 16 b, coefficient shielding (uses shadow unit 1615A-D and other 20 unlabelled shadow unit are described) can serve to indicate that at the z word to terminate towards DC coefficient 1605 The AC conversion coefficient of (or alternatively to omit) is included during type sequential scan.
Figure 17 shows the quick IC code book filter according to an embodiment with frequency spectrum more new encoder and renewable code book Ripple device encoder 1700.IC code book wave filter encoder 1700 is similar to IC code book wave filter discussed above in many aspects and compiles Code device 1400.But, unlike IC code book wave filter encoder 1400, IC code book wave filter encoder 1700 can be by more New IC filter coefficient (icfcu) stores in the codebook, as follows.Controller 1755 is via storage signal (str) instruction code book The IC filter coefficient (icfcu) updated with storage should be updated.Storage signal (str) by switch 1758 use to determine is The no IC filter coefficient (icfcu) by renewal stores in the codebook.Storage signal (str) is also coded by entropy device 1718 and is encoded to Bit stream 1728.In certain embodiments, controller 1755 also can indicate that code book reset signal (not shown), and its instruction should Abandoning the renewal of storage, code book returns to its canonical form (such as, when scene changes).
IC code book wave filter encoder 1700 also resides in from the different of IC code book wave filter encoder 1400: conversion coefficient screen Covering (seeing Figure 16 b, discussed above) is (not shield before (shielding 1758) and Transform operations after map function 1760) use.
Figure 18 shows the IC code book wave filter decoding corresponding with IC code book wave filter encoder 1700 discussed above Device 1800.Wave filter decoder 1800 is similar to IC code book wave filter decoder 1500 discussed above in many aspects.But, Unlike IC code book wave filter decoder 1500, the IC wave filter system that IC code book wave filter decoder 1800 can will update Number (icfcu) stores in the codebook, as follows.Storage signal (str) from bit stream 1801 is solved by entropy decoder 1803 Code.Storage signal (str) is switched on and off 1758 uses to determine whether the IC filter coefficient (icfcu) updated is stored in code book In.In certain embodiments, reset signal (not shown) is also decoded by entropy decoder 1803, reset signal instruction storage Renewal should be dropped, and code book returns to its canonical form.
Figure 19 shows the example bitstream grammer 1900 of IC code book wave filter encoder, and in this article Carry out various description.Bitstream syntax 1900 starts from the header cicf_hdr 1901 of 2 bits.The pict_only mark of one bit Note 1902 indicates whether to specify single IC wave filter for whole picture.Optional 3 bit numpt fields 1903 indicate subregion Quantity (up to eight).Optional 1 bit map_sent labelling 1904 indicates whether that sending subregion maps so that where instruction answers With or do not apply wave filter, or alternatively switch between multiple wave filter.Optional n-bit field 1905 includes subregion Map (if present).
The fsnt labelling 1906 of one bit indicates whether IC wave filter includes in the bitstream.If labelling 1906 is set , then field fsz 1907 of optional 1 or 2 bits indicate included IC wave filter wave filter size (such as, 5 × 5,7 × 7 or 9 × 9) it is, the code word of institute's selecting filter afterwards.Code word includes that the icfci field 1908 of 8 bits is (to IC wave filter The index of code book) and 2 bits (separable filter) or the flt_ornt field 1909 of 3 bits (inseparable wave filter), Its instruction should apply to the wave filter modification of index coefficient indicated by icfci 1908 or converts (such as, for separable Wave filter is in four rotations, or is in four rotations and four mirror images for inseparable wave filter). Whether optional 1 bit updt labelling 1910 indication bit stream includes the difference for the coefficient indicated by index icfci 1908 Update.The updt_dat field 1911 of optional s bit includes to be applied in indexing the coefficient indicated by icfci 1908 Difference update.When being indicated by pict_only labelling 1902, field 1906-1911 repeats numpt 1903 times.
Figure 20 a shows the picture in video flowing or the sequence of frame 2001-2008, including I-frame 2001,2008, P-frame 2004,2006 and B-frame 2002-2003,2005,2007.Figure 20 b shows the picture in video flowing or frame 2001-2008's Sequence, including: I-frame 2001,2008, it is divided into three exemplary partitions (numpt=3);P-frame 2004,2006, its quilt It is divided into three exemplary partitions (numpt=3);And B-frame 2002-2003,2005,2007, it is divided into two examples Property subregion (numpt=2).Table 1 indicates the various combinations of the number of partitions (numpt) according to picture/mb-type to IC code book wave filter Carry out the exemplary encoding bit overhead encoded.(the exemplary encoding bit overhead shown in table 1 eliminates space partition zone and maps (such as region, sheet, tile or tile segmentation, picture binary tree or quaternary tree, tile binary tree or quaternary tree and/or merging Block) overhead bit).
Table 1
Figure 21 shows and uses code book to be encoded in bit stream by the IC wave filter of the picture of video according to an embodiment Routine 2100.In square frame 2105, routine 2100 obtains code book, and described code book includes multiple IC filter coefficient set, its It is suitable for the compensation for loss and damage wave filter configuring Video Decoder, to process, there is various characteristic (such as high-contrast, low contrast Deng) the frame of encoded video.In one embodiment, code book can include 256 filter coefficient set, and they are according to generation Table coefficient is ranked up (such as, being ranked up) according to center coefficient.Other embodiments can include carrying out according to different criterions The smaller or greater code book of sequence.
Starting in the square frame 2110 started the cycle over, routine 2100 processes each picture or the frame of video.At square frame 2115 In, routine 2100 analyzes photo current.Such as, in one embodiment, routine 2100 may determine that one or more image is special Property, such as contrast level and/or compared with picture before photo current indicate whether scene changes.
In square frame 2120, routine 2100 determines one or more parts of photo current.Such as, implement according to each Example, Figure 10 a-f discussed above shows the various sample portion that picture can be divided into.In certain embodiments, may be used Number of partitions determined by routine 2100 is determined with the picture/mb-type based in part on photo current.(see, e.g. figure 20b, discussed above.)
In square frame 2125, one or more Picture section designators are encoded in bit stream by routine 2100.Such as, exist In one embodiment, routine 2100 can in field 1902-1905 shown in Figure 19 and discussed above or Multiple encode.
Starting in the square frame 2130 started the cycle over, it is every that routine 2100 processes in one or more Picture section determined One.In square frame 2150, routine 2100 selects IC coefficient sets to improve in this IC coefficient sets adaptively from code book Compensation for loss and damage wave filter for configuring decoder reproduces photo current part with Video Decoder during process photo current part Informativeness.Figure 12 discussed above, 14 and 17 show the various methods being adaptive selected IC coefficient sets from code book. Such as, in one embodiment, be adaptive selected IC coefficient sets can include following in some or all: to uncoded Picture encodes, and is decoded coded picture in local decoder loop;According to the picture decoded with do not compile Code picture calculates the filter coefficient set of target IC, and selects substantially to accord with the filter coefficient set of target IC from code book The IC filter coefficient set closed.In certain embodiments, select the IC filter coefficient substantially met with object set permissible Including: according to the coefficient sets representing coefficient (such as center coefficient) selection approximate match;And identify with approximate match be Manifold closes the comparable group of neighbouring IC wave filter.IC wave filter (including the zero or more modification of each) in comparable group with After can be determined by error estimation (such as MSE) and be scored, wherein, select the IC filter coefficient set that scoring is minimum.
In square frame 2155, routine 2100 determines the wave filter designator of selected IC coefficient sets, such as, arrive the rope of code book Draw.In square frame 2160, wave filter designator is encoded in bit stream by routine 2100, so that decoder can be from decoder Code book copy in position selected IC coefficient sets.
In decision block 2165, routine 2100 determine whether send for selected IC coefficient sets renewal with deposit Storage selected IC coefficient sets in the codebook compares the informativeness improving Video Decoder reproduction photo current part further. In certain embodiments, can not use update mechanism, in this case, the determination in square frame 2156 can be negated always 's.In the embodiment using update mechanism, routine 2100 may determine that whether the IC coefficient of renewal can change replication fidelity It is apt to enough degree to prove that it is proper for sending the overhead bit updated.If it is determined that update, then in subroutine side In frame 2200 (seeing Figure 22, be discussed below), adjusted value is determined and is encoded to the IC coefficient at decoder selected by renewal Set.In certain embodiments, adjusted value can include the conversion coefficient quantified, its represent the filter coefficient that calculated with Difference between the approximate match found from code book.
In the square frame 2190 of end loop, routine 2100 be circulated back to square frame 2130 with process next Picture section (as If fruit exists).In the square frame 2195 of end loop, routine 2100 be circulated back to square frame 2110 with process next picture (as If fruit exists).Once have been processed by all of picture, then routine 2100 terminates in square frame 2199.
Figure 22 shows that the IC filter coefficient set according to an embodiment updates subroutine 2200.At square frame 2205 In, subroutine 2200 determines for adjusting the adjusted value set of the IC filter coefficient set of storage in code book.Such as, at one In embodiment, adjusted value can include the difference coefficient value between set and the filter coefficient set of target IC that code book stores.
In square frame 2210, adjusted value is encoded in bit stream by subroutine 2200 so that decoder is obtained in that described Adjusted value.In certain embodiments, adjusted value is carried out coding and can include calculating the frequency domain representation of adjusted value, and according to inverse z Font scanning sequency carries out entropy code to frequency domain representation, may only include AC coefficient shielding subset (see, e.g. Figure 16 b, Discussed above).
In square frame 2215, adjustment designator is encoded in bit stream by subroutine 2200, and described adjustment designator instructs Video Decoder configuration decoder compensation for loss and damage wave filter with process photo current current portions before according to encoded Adjusted value adjust in code book the IC filter coefficient set of storage.(see, e.g. shown in Figure 19 and discussed above Bit field 1910-1911.)
In decision block 2220, subroutine 2200 determines whether the IC filter coefficient set of adjustment is stored in code book In enable to the IC filter coefficient reusing renewal without again sending adjusted value.If it is, at square frame In 2225, the IC filter coefficient set of adjustment is stored in the local code book copy of encoder by subroutine 2200, and In square frame 2230, storage indicator is encoded to bit stream and adjusts in designator by subroutine 2200, instructs Video Decoder to adjust Whole coefficient value is stored in the code book copy of decoder.In certain embodiments, memory mechanism can not be used, in this feelings In condition, the determination in square frame 2220 can be negated always.In the embodiment using memory mechanism, routine 2100 may determine that Whether the IC coefficient adjusted may be reused afterwards.Subroutine 2200 terminates in square frame 2299.
Figure 23 shows and uses one or more code books to be encoded by the IC wave filter of the picture of video according to an embodiment Routine 2300 in bit stream.In square frame 2305, routine 2300 obtains one or more code book, and each code book includes giving Multiple coefficient sets of the IC wave filter of size.In one embodiment, each code book can include carrying out according to representing coefficient 256 filter coefficient set of sequence (such as, being ranked up according to center coefficient).Other embodiments can include according to not The smaller or greater code book being ranked up with criterion.In one embodiment, one or more code books can include that size is 3 The coefficient of the IC wave filter of × 3,5 × 5,7 × 7,9 × 9,11 × 11 etc..
Starting in the square frame 2310 started the cycle over, routine 2300 processes each picture or the frame of video.At square frame In 2315, routine 2300 analyzes photo current.Such as, in one embodiment, routine 2300 may determine that one or more figure As characteristic, such as contrast level and/or compared with picture before photo current indicate whether scene changes.
In square frame 2320, routine 2300 determines one or more parts of photo current.Such as, implement according to each Example, Figure 10 a-f discussed above shows the various sample portion that picture can be divided into.In certain embodiments, may be used Partial amt determined by routine 2300 is determined with the picture/mb-type based in part on photo current.(see, e.g. figure 20b, discussed above.)
In square frame 2325, one or more Picture section designators are encoded in bit stream by routine 2300.Such as, exist In one embodiment, routine 2300 can be to one or more field 1902-1905 shown in Figure 19 and discussed above Encode.
In square frame 2326, routine 2300 determines the wave filter size of photo current.In square frame 2327, routine 2300 will Determined by wave filter size coding in bit stream (see for example the field 1907 in Figure 19, discussed above.) such as, When photo current is " low contrast ", it may be determined that bigger wave filter size (such as, 9 × 9).On the other hand, at input figure When sheet is " high-contrast ", if coding quality is high, may determine that less wave filter size (such as, 5 × 5), or If coding quality waiting until low, may determine that medium wave filter size (such as, 7 × 7).For medium-contrast Scene for, may determine that medium wave filter size (such as, 7 × 7) when coding quality is enough, or at coding quality Big wave filter size (such as, 9 × 9) is may determine that time low.In certain embodiments, wave filter size information be at picture or It is transmitted on the basis of frame.In other embodiments, wave filter size information can be carried out for each part of picture Determine and send (not shown).
In square frame 2328, routine 2300 selects the code book with the coefficient sets of the IC wave filter of selected size.Opening Begin the square frame 2330 of circulation starts, routine 2300 process one or more determined by each in Picture section.
In square frame 2350, routine 2300 is adaptive selected IC coefficient sets to improve in this IC coefficient set from code book Share Video Decoder when the compensation for loss and damage wave filter of configuration decoder is to process photo current part and reproduce photo current portion The informativeness divided.
In square frame 2355, routine 2300 determines the wave filter designator of selected IC coefficient sets, such as, arrive code book Index.In square frame 2360, routine 2300 is encoded in bit stream by wave filter designator so that decoder can be from decoding The IC coefficient sets selected by code book replica location of device.
In decision block 2365, routine 2300 determine whether send for selected IC coefficient sets renewal with deposit Storage selected IC coefficient sets in the codebook compares the informativeness improving Video Decoder reproduction photo current part further. In certain embodiments, can not use update mechanism, in this case, the determination in square frame 2356 can be negated always 's.In the embodiment using update mechanism, routine 2300 may determine that whether the IC coefficient of renewal can change replication fidelity It is apt to enough degree to prove that it is proper for sending the overhead bit updated.If it is determined that update, then in subroutine side In frame 2200 (seeing Figure 22, discussed above), adjusted value is determined and is encoded to the IC coefficient at decoder selected by renewal Set.
In the square frame 2390 of end loop, routine 2300 be circulated back to square frame 2330 with process next Picture section (as If fruit exists).In the square frame 2395 of end loop, routine 2300 be circulated back to square frame 2310 with process next picture (as If fruit exists).Once having been processed by all pictures, routine 2300 just terminates in square frame 2399.
Figure 24 shows and uses one or more code books IC to the video pictures from bit stream according to an embodiment The routine 2400 that wave filter is decoded.In square frame 2403, routine 2400 obtains one or more code book, and each code book includes Have to multiple coefficient sets of the IC wave filter of sizing.These code books be can be used for coding routine discussed above 220 and/ Or the copy of the code book of 2400.
In square frame 2405, routine 2400 obtains encoded bit stream, and it includes that the multiple pictures with video or frame are relative The coded data answered.Starting in the square frame 2408 started the cycle over, routine 2400 processes coding each figure in the bitstream Sheet.
In decision block 2410, routine 2400 determines according to the one or more reset signals being encoded in bit stream Whether one or more code books are reset.Such as, in one embodiment, encoder can instruct routine 2400 in scene During change or resetting code book At All Other Times.If the reset signal instruction in bit stream should reset one or more code book (renewal before i.e., arbitrarily is dropped), then in square frame 2413, indicated code book returns to their initial condition.
In square frame 2415, current encoded picture is decoded by routine 2400.In square frame 2418, routine 2400 is right One or more Picture section designators from bit stream are decoded.Such as, in one embodiment, routine 2400 is permissible One or more in field 1902-1905 shown in Figure 19 and discussed above are decoded.
Starting in the square frame 2420 started the cycle over, it is every that routine 2400 processes in the Picture section of one or more instruction One.In square frame 2423, the one or more wave filter designators from bit stream are decoded by routine 2400.Such as, In one embodiment, routine 2400 can be in field 1906-1909 shown in Figure 19 and discussed above Or multiple be decoded.
In square frame 2425, routine 2400 is according to one or more (the such as fields in the wave filter designator decoded 1907) in the one or more code books obtained in square frame 2403 is selected.In certain embodiments, iff depositing At the single code book obtained in square frame 2403, then that can ignore in decoded wave filter designator is one or more, And routine 2400 can simply choose single code book.In square frame 2426, routine 2400 select from selected code book by The IC filter coefficient set that one or more decoded wave filter designators (such as field 1908) indicate.Implement at some In example, IC filter coefficient set is selected to include: to calculate modification or conversion (such as, the rotation of IC filter coefficient set Turn or mirror image), such as can be indicated by field 1909.
In decision block 2428, routine 2400 determines that bit stream has instructed routine 2400 to adjust or updated selected IC filter coefficient set.Such as, in one embodiment, such guidance can send via field 1910.If Do not send renewal to instruct, then in square frame 2430, routine 2400 is according to selected IC coefficient set incompatible configuration IC wave filter. On the other hand, instruct if having sent renewal, then, in square frame 2433, routine 2400 is to the adjusted value from bit stream (such as, send via field 1911) is decoded, and updates selected IC filtering according to adjusted value in square frame 2435 Device coefficient sets.In square frame 2430, routine 2400 is according to IC filter coefficient set incompatible configuration IC filter that is that update or that adjust Ripple device.
In decision block 2440, routine 2400 determines that bit stream has instructed 2400 by IC that is that update or that adjust Filter coefficient set is stored in the code book copy of decoder, so that the IC filter coefficient updated can be reused.As Fruit has been received that such guidance, then, in square frame 2443, routine 2400 is by IC filter coefficient that is that update or that adjust Set is stored in the current code book copy of decoder.
In square frame 2445, routine 2400 uses the IC wave filter configured to process photo current part to improve decoding Picture section reproduce the informativeness of appropriate section of uncoded original image.
In the square frame 2448 of end loop, routine 2400 be circulated back to square frame 2420 with process next Picture section (as If fruit exists).In the square frame 2450 of end loop, routine 2400 be circulated back to square frame 2408 with process next picture (as If fruit exists).The most all of picture is processed, and routine 2400 just terminates in square frame 2499.
Although specific embodiment is had been illustrated and described herein, but those of ordinary skill in the art it will be clear that That can be replaced with and/or equivalence realization is to replace specific embodiment that is shown and that describe, without deviating from present disclosure Scope.The application is intended to arbitrarily adjusting or modification of embodiment discussed herein.

Claims (48)

1. the method that video encoder apparatus realizes, for the compensation for loss and damage filter coefficient value collection to encoded frame of video Conjunction encodes, and described method includes:
The code book including multiple compensation for loss and damage filter coefficient value set, the plurality of damage is obtained by described video encoder apparatus Evil compensating filter sets of coefficient values is suitable for configuring the compensation for loss and damage wave filter of Video Decoder with to having multifrequency nature The frame of encoded video processes;
During encoding to produce encoded bit stream to uncoded frame of video, described video encoder apparatus is adaptively Compensation for loss and damage filter coefficient value set is selected, to improve at described compensation for loss and damage filter coefficient value set from described code book Decode with described video when processing at least partially to described encoded frame for configuring described compensation for loss and damage wave filter Device reproduces at least one of informativeness of described uncoded frame of video;And
Being included in by one designator in described encoded bit stream produced by described video encoder apparatus, described designator makes Obtaining described Video Decoder can select selected compensation for loss and damage to filter from the copy of the described Video Decoder of described code book Device sets of coefficient values.
Described compensation for loss and damage wave filter system is selected the most adaptively from described code book Numerical value set includes:
Described uncoded frame of video is encoded;
Encoded frame is decoded in local decoder loop by described video encoder apparatus;And
Target lesion compensating filter sets of coefficient values is calculated according to decoded frame and described uncoded frame.
3. method as claimed in claim 2, wherein, selects described compensation for loss and damage filter coefficient value set from described code book Including: from described code book, select the compensation for loss and damage being substantially consistent with described target lesion compensating filter sets of coefficient values Filter coefficient value set.
4. method as claimed in claim 3, wherein, select from described code book substantially with described target lesion compensation filter The compensation for loss and damage filter coefficient value set that device sets of coefficient values is consistent includes:
According to one or more values representing coefficient of described target lesion compensating filter sets of coefficient values, from described code book Select the compensation for loss and damage filter coefficient value set of approximate match;
Select comparable group, described comparable group include the compensation for loss and damage filter coefficient value set of described approximate match and one or Multiple neighbouring compensation for loss and damage filter coefficient value sets;
Determine between described target lesion compensating filter sets of coefficient values and the set including described comparable group respectively is multiple Comparison measuring;And
Multiple comparison measurings determined by according to, select from described code book substantially with described target lesion compensating filter system The compensation for loss and damage filter coefficient value set that numerical value set is consistent.
5. method as claimed in claim 4, wherein, determines that the plurality of comparison measuring includes: determine described mesh respectively respectively Multiple error estimations between mark compensation for loss and damage filter coefficient value set and described comparable group.
6. method as claimed in claim 4, wherein, described code book is to represent coefficient according to the one or more to be ranked up Or index, to contribute to selecting the set of described approximate match.
7. method as claimed in claim 4, wherein, the one or more represents coefficient and includes the center of Wiener wave filter Coefficient.
8. method as claimed in claim 4, wherein, described comparable group also includes: the compensation for loss and damage filtering of described approximate match At least one in device sets of coefficient values and the one or more neighbouring compensation for loss and damage filter coefficient value set is at least One modification calculated.
9. the method for claim 1, also includes: during encoding described uncoded frame of video, calculates selected The modification of compensation for loss and damage filter coefficient value set.
10. method as claimed in claim 9, wherein, the modification bag of the compensation for loss and damage filter coefficient value set selected by calculating Include following in one: the compensation for loss and damage filter coefficient value set selected by rotation, and reversion selected by compensation for loss and damage wave filter Sets of coefficient values.
11. methods as claimed in claim 9, also include: the second designator is included in described video encoder apparatus and is produced In raw described encoded bit stream, described second designator makes described Video Decoder can calculate selected compensation for loss and damage The modification calculated of filter coefficient value set.
12. the method for claim 1, also include: during encoding described uncoded frame of video, adaptively The second compensation for loss and damage filter coefficient value set is selected, to improve in described second compensation for loss and damage wave filter system from described code book Set of values is share and is regarded described in time configuring described compensation for loss and damage wave filter and process with the Part II to described encoded frame The informativeness of the Part II of the described uncoded frame of video of decoder reproduction frequently.
13. methods as claimed in claim 12, also include: the second designator is included in described video encoder apparatus and is produced In raw described encoded bit stream, described second designator makes described Video Decoder be capable of identify that described encoded frame Described part and described Part II.
14. the method for claim 1, also include:
During described uncoded frame of video is encoded:
Determining multiple adjusted value, the plurality of adjusted value is for adjusting the phase in selected compensation for loss and damage filter coefficient value set The multiple values answered, to improve the described at least one of loyalty of the described Video Decoder described uncoded frame of video of reproduction further Solidity;And
According to adjusted value determined by multiple update in described code book selected by compensation for loss and damage filter coefficient value set;With And
More new data and renewal instruction are included in described encoded bit stream, described more new data and the plurality of adjusted value Corresponding, the described instruction described Video Decoder of guidance that updates updates the described video of described code book according to described more new data The copy of decoder.
15. methods as claimed in claim 14, also include: follow-up second uncoded frame of video is encoded during, will Resetting instruction to be included in described encoded bit stream, described replacement before the instruction described Video Decoder of guidance abandons is instructed The renewal of copy of the described Video Decoder for described code book.
16. 1 kinds for the device encoding the compensation for loss and damage filter coefficient value set of encoded frame of video, described dress Put and include:
For being obtained the module of the code book including multiple compensation for loss and damage filter coefficient value set, institute by video encoder apparatus State multiple compensation for loss and damage filter coefficient value set to be suitable for configuring the compensation for loss and damage wave filter of Video Decoder with many to having The frame of the encoded video planting characteristic processes;
For during encoding to produce encoded bit stream to uncoded frame of video, by described video encoder apparatus Compensation for loss and damage filter coefficient value set is selected adaptively, to improve in described compensation for loss and damage wave filter system from described code book Set of values share in configure described compensation for loss and damage wave filter with to described encoded frame when processing at least partially described in Video Decoder reproduces the module of at least one of informativeness of described uncoded frame of video;And
For a designator being included in the module in described encoded bit stream produced by described video encoder apparatus, institute State designator make described Video Decoder can select from the copy of the described Video Decoder of described code book selected by damage Evil compensating filter sets of coefficient values.
17. 1 kinds of video coding apparatus, including:
Compensation for loss and damage wave filter code book look-up table and address logic, for uncoded frame of video is encoded with produce compile During code bit stream, from code book, select compensation for loss and damage filter coefficient value set adaptively, mend in described infringement to improve Repay filter coefficient value set for configuring at least some of with to encoded frame of the compensation for loss and damage wave filter of Video Decoder When processing, described Video Decoder reproduces at least one of informativeness of described uncoded frame of video, wherein, described code Originally include that multiple compensation for loss and damage filter coefficient value set, the plurality of compensation for loss and damage filter coefficient value set are suitable for configuration The described compensation for loss and damage wave filter of described Video Decoder processes with the frame to the encoded video with multifrequency nature;With And
Entropy coder, for being included in described encoded bit stream by a designator, described designator makes described video solution Code device can select selected compensation for loss and damage filter coefficient value set from the copy of the described Video Decoder of described code book.
The method that 18. 1 kinds of video encoder apparatus realize, for the compensation for loss and damage filter coefficient value to encoded frame of video Set encodes, and described method includes:
Being obtained at least two code book by described video encoder apparatus, the first code book in described at least two code book includes first Multiple compensation for loss and damage filter coefficient value sets, described more than first compensation for loss and damage filter coefficient value set is suitable for configuration tool There is the Video Decoder compensation for loss and damage wave filter of the first wave filter size to regard having the encoded of at least the first picture characteristics The frame of frequency processes, and the second code book in described at least two code book includes more than second compensation for loss and damage filter coefficient value collection Closing, described more than second compensation for loss and damage filter coefficient value set is suitable for configuring the video decoding with the second wave filter size Device compensation for loss and damage wave filter processes with the frame to the encoded video with at least the second picture characteristics, wherein, and described Two wave filter sizes are less than described first wave filter size;
During encoding to produce encoded bit stream to uncoded frame of video, described video encoder apparatus:
Determine the picture characteristics of described uncoded frame;
Picture characteristics determined by being based at least partially on is to select in described at least two code book;And
From selected code book, select compensation for loss and damage filter coefficient value set adaptively, filter in described compensation for loss and damage to improve Described video solution when ripple device sets of coefficient values is used for configuring described compensation for loss and damage wave filter to process described encoded frame Code device reproduces the informativeness of described uncoded frame of video;And
First designator and the second designator are included in described encoded bit stream produced by described video encoder apparatus In, described first designator makes described Video Decoder can select selected code book, institute from described at least two code book Stating the second designator makes described Video Decoder can select selected compensation for loss and damage filter coefficient from selected code book Value set.
19. methods as claimed in claim 18, wherein, be based at least partially on determined by picture characteristics select described in extremely One in few two code books includes:
Target filter size is determined according to picture characteristics determined by described uncoded frame;And
Determined by being based at least partially on, target filter size is to select code book.
20. methods as claimed in claim 19, wherein it is determined that the picture characteristics of described uncoded frame comprise determining that described in not The contrast level of coded frame.
21. methods as claimed in claim 20, wherein, determine institute according to picture characteristics determined by described uncoded frame State target filter size to include: determine described target filter according to contrast level determined by described uncoded frame Size.
22. methods as claimed in claim 21, wherein, determine according to contrast level determined by described uncoded frame Described target filter size includes: determine bigger filtering when contrast level determined by described uncoded frame is relatively low Device size, and less wave filter size is determined when contrast level determined by described uncoded frame is higher.
23. methods as claimed in claim 19, wherein, select a ground the most at least part of in described at least two code book In the encoding quality metrics being associated with described coding.
24. methods as claimed in claim 23, wherein, determine institute according to picture characteristics determined by described uncoded frame State target filter size to include: determine described target filter size according to described encoding quality metrics.
According to described encoding quality metrics, 25. methods as claimed in claim 24, wherein, determine that described target filter is big Little include: determine bigger wave filter size when described coding quality is relatively low, and true when described encoding quality metrics is higher Fixed less wave filter size.
26. methods as claimed in claim 18, wherein, described more than first compensation for loss and damage filter coefficient value set includes using The sets of coefficient values of a compensation for loss and damage wave filter selected in the compensation for loss and damage wave filter of 5 × 5,7 × 7 and 9 × 9, and Wherein, described more than second compensation for loss and damage filter coefficient value set includes that the compensation for loss and damage for 5 × 5,7 × 7 and 9 × 9 is filtered The sets of coefficient values of a different compensation for loss and damage wave filter selected in ripple device.
27. 1 kinds for the device encoding the compensation for loss and damage filter coefficient value set of encoded frame of video, described dress Put and include:
For being obtained the module of at least two code book, the first code book in described at least two code book by video encoder apparatus Including more than first compensation for loss and damage filter coefficient value set, described more than first compensation for loss and damage filter coefficient value set is suitable for There is the Video Decoder compensation for loss and damage wave filter of the first wave filter size with to having at least the first picture characteristics in configuration The frame of encoded video processes, and the second code book in described at least two code book includes more than second compensation for loss and damage wave filter Sets of coefficient values, described more than second compensation for loss and damage filter coefficient value set is suitable for configuring and has the second wave filter size Video Decoder compensation for loss and damage wave filter processes with the frame to the encoded video with at least the second picture characteristics, its In, described second wave filter size is less than described first wave filter size;
During encoding to produce encoded bit stream to uncoded frame of video:
For being determined the module of the picture characteristics of described uncoded frame by described video encoder apparatus;
Determined by be based at least partially on picture characteristics by described video encoder apparatus select described at least two The module of in individual code book;And
For being selected compensation for loss and damage filter coefficient value adaptively from selected code book by described video encoder apparatus Set, with improve described compensation for loss and damage filter coefficient value set be used for configuring described compensation for loss and damage wave filter with to described When coded frame processes, Video Decoder reproduces the module of the informativeness of described uncoded frame of video;And
For the first designator and the second designator being included in described encoded ratio produced by described video encoder apparatus Module in special stream, it is selected that described first designator makes described Video Decoder can select from described at least two code book Code book, described second designator makes described Video Decoder selected compensation for loss and damage can be selected from selected code book to filter Ripple device sets of coefficient values.
28. 1 kinds of video coding apparatus, look into including compensation for loss and damage wave filter codebook search processor, compensation for loss and damage wave filter code book Look for table and address logic and entropy coder, during encoding to produce encoded bit stream to uncoded frame of video:
Described compensation for loss and damage wave filter codebook search processor is for determining the picture characteristics of described uncoded frame, and at least partly Ground based on determined by picture characteristics select at least two code book, wherein, first in described at least two code book Code book includes more than first compensation for loss and damage filter coefficient value set, described more than first compensation for loss and damage filter coefficient value set It is suitable for configuring the Video Decoder compensation for loss and damage wave filter with the first wave filter size with special to having at least the first image Property the frame of encoded video process, the second code book in described at least two code book includes more than second compensation for loss and damage filter Ripple device sets of coefficient values, described more than second compensation for loss and damage filter coefficient value set is suitable for configuration, and to have the second wave filter big Little Video Decoder compensation for loss and damage wave filter processes with the frame to the encoded video with at least the second picture characteristics, Wherein, described second wave filter size is less than described first wave filter size;
Described compensation for loss and damage wave filter code book look-up table and address logic select infringement from selected code book for adaptively Compensating filter sets of coefficient values, is used for configuring described compensation for loss and damage at described compensation for loss and damage filter coefficient value set to improve When wave filter is to process described encoded frame, described Video Decoder reproduces the informativeness of described uncoded frame of video;And And
Described entropy coder is used for being included in described encoded bit stream the first designator and the second designator, and described first Designator makes described Video Decoder can select selected code book, described second designator from described at least two code book Make described Video Decoder can select selected compensation for loss and damage filter coefficient value set from selected code book.
The method that 29. 1 kinds of video encoder apparatus realize, for the compensation for loss and damage filter coefficient value to encoded frame of video Set encodes, and described method includes:
The code book including multiple compensation for loss and damage filter coefficient value set, the plurality of damage is obtained by described video encoder apparatus Evil compensating filter sets of coefficient values is suitable for configuring the compensation for loss and damage wave filter of Video Decoder with to having multifrequency nature The frame of encoded video processes;
During encoding to produce encoded bit stream to uncoded frame of video, described video encoder apparatus:
From described code book, select compensation for loss and damage filter coefficient value set adaptively, filter in described compensation for loss and damage to improve Device sets of coefficient values is for configuring described compensation for loss and damage wave filter with to when processing at least partially of described encoded frame Described Video Decoder reproduces at least one of informativeness of described uncoded frame of video;
Determining multiple adjusted value, the plurality of adjusted value is for adjusting the phase in selected compensation for loss and damage filter coefficient value set The multiple values answered, to improve the described at least one of loyalty of the described Video Decoder described uncoded frame of video of reproduction further Solidity;And
The plurality of adjusted value is encoded;And
It is included in described video encoder apparatus is produced selecting designator, encoded multiple adjusted values and adjustment designator Described encoded bit stream in, described selection designator makes the described Video Decoder can be from the described video of described code book Selecting selected compensation for loss and damage filter coefficient value set in the copy of decoder, described adjustment designator instructs described video solution Code device configure described compensation for loss and damage wave filter with process described encoded frame described at least some of before according to described Multiple adjusted values of coding adjust selected compensation for loss and damage filter coefficient value set.
30. methods as claimed in claim 29, wherein, select described compensation for loss and damage wave filter adaptively from described code book Sets of coefficient values includes:
Described uncoded frame of video is encoded;
Encoded frame is decoded in local decoder loop by described video encoder apparatus;And
Target lesion compensating filter sets of coefficient values is calculated according to decoded frame and described uncoded frame.
31. methods as claimed in claim 30, wherein, the plurality of adjusted value is to determine according to multiple difference coefficient values , the plurality of difference coefficient value represents that described target lesion compensating filter sets of coefficient values filters with selected compensation for loss and damage Difference between device sets of coefficient values.
32. methods as claimed in claim 31, wherein, carry out coding to the plurality of adjusted value and include: calculate the plurality of The frequency domain representation of difference coefficient value.
33. methods as claimed in claim 32, wherein, encode the plurality of adjusted value and also include: according to described The z font scanning sequency that the DC coefficient of the frequency domain representation of multiple difference coefficient values terminates, the frequency to the plurality of difference coefficient value Domain representation carries out entropy code.
34. methods as claimed in claim 32, wherein, encode the plurality of adjusted value and also include: shield described many Multiple AC coefficients of the frequency domain representation of individual difference coefficient value.
35. 1 kinds for the device encoding the compensation for loss and damage filter coefficient value set of encoded frame of video, described dress Put and include:
For being obtained the module of the code book including multiple compensation for loss and damage filter coefficient value set, institute by video encoder apparatus State multiple compensation for loss and damage filter coefficient value set to be suitable for configuring the compensation for loss and damage wave filter of Video Decoder with many to having The frame of the encoded video planting characteristic processes;
During encoding to produce encoded bit stream to uncoded frame of video:
For being selected compensation for loss and damage filter coefficient value collection adaptively from described code book by described video encoder apparatus Close, be used for configuring described compensation for loss and damage wave filter to compile described at described compensation for loss and damage filter coefficient value set to improve Described Video Decoder when processing at least partially of code frame reproduces at least one of loyalty of described uncoded frame of video The module of solidity;
For being determined the module of multiple adjusted value by described video encoder apparatus, the plurality of adjusted value is used for adjusting selected Compensation for loss and damage filter coefficient value set in corresponding multiple values, reproduce described improving further described Video Decoder The module of the described at least one of informativeness of uncoded frame of video;And
For the module the plurality of adjusted value encoded by described video encoder apparatus;And
For designator, encoded multiple adjusted values and adjustment designator being selected to be included in described video encoder apparatus institute Module in the described encoded bit stream produced, described selection designator makes the described Video Decoder can be from described code book Described Video Decoder copy in select selected compensation for loss and damage filter coefficient value set, described adjustment designator instructs Described Video Decoder configure described compensation for loss and damage wave filter with process described encoded frame described at least some of before Selected compensation for loss and damage filter coefficient value set is adjusted according to described encoded multiple adjusted values.
36. 1 kinds of video coding apparatus, including compensation for loss and damage wave filter code book look-up table and address logic, difference engine and entropy code Device, during encoding to produce encoded bit stream to uncoded frame of video:
Described compensation for loss and damage wave filter code book look-up table and address logic select compensation for loss and damage filter from code book for adaptively Ripple device sets of coefficient values, mends for the infringement configuring Video Decoder at described compensation for loss and damage filter coefficient value set to improve Repay wave filter and reproduce described uncoded frame of video with described Video Decoder when processing at least partially to encoded frame At least one of informativeness, wherein, described code book includes multiple compensation for loss and damage filter coefficient value set, the plurality of damage Evil compensating filter sets of coefficient values is suitable for configuring the described compensation for loss and damage wave filter of described Video Decoder with many to having The frame of the encoded video planting characteristic processes;
Described difference engine is used for determining multiple adjusted value, and the plurality of adjusted value is for adjusting selected compensation for loss and damage wave filter system Corresponding multiple values in numerical value set, to improve further described in the described Video Decoder described uncoded frame of video of reproduction At least one of informativeness;And
Described entropy coder is for encoding the plurality of adjusted value and will select designator, encoded multiple adjusted values Being included in described encoded bit stream with adjusting designator, described selection designator makes the described Video Decoder can be from institute Selecting selected compensation for loss and damage filter coefficient value set in the copy of the described Video Decoder stating code book, described adjustment indicates Symbol instructs described Video Decoder configuring described compensation for loss and damage wave filter with described in the described encoded frame of process at least one / front adjust selected compensation for loss and damage filter coefficient value set according to described encoded multiple adjusted values.
The method that 37. 1 kinds of video decoder apparatus realize, for the compensation for loss and damage filter coefficient value to encoded frame of video Set is decoded, and described method includes:
The code book including multiple compensation for loss and damage filter coefficient value set, the plurality of damage is obtained by described video decoder apparatus Evil compensating filter sets of coefficient values is suitable for configuring compensation for loss and damage wave filter with to the decoded video with multifrequency nature Frame processes;
Obtaining encoded bit stream, described encoded bit stream includes described encoded frame of video and selects designator, described choosing Selecting designator makes described Video Decoder can select compensation for loss and damage filter coefficient value set from described code book;
Described encoded frame of video is being decoded period, and described video decoder apparatus selects by described from described code book Select the compensation for loss and damage filter coefficient value set of designator instruction;
Described compensation for loss and damage is configured according to selected compensation for loss and damage filter coefficient value set by described video decoder apparatus Wave filter;
Used the compensation for loss and damage wave filter configured to process described decoded video frame extremely by described video decoder apparatus A few part, to improve the informativeness of the described decoded video frame corresponding uncoded source video frame of reproduction;And
Handled decoded video frame is supplied to the display being associated with described video decoder apparatus.
38. methods as claimed in claim 37, wherein, described encoded bit stream also includes the second selection designator and part Identification indicator, described second selects designator to make described Video Decoder that the second infringement can be selected from described code book to mend Repaying filter coefficient value set, described portion identification designator makes described Video Decoder be capable of identify that described decoded frame Described part and Part II.
39. methods as claimed in claim 38, also include:
Described encoded frame of video is being decoded period, and described video decoder apparatus selects by described from described code book Second the second compensation for loss and damage filter coefficient value set selecting designator instruction;
Described compensation for loss and damage wave filter is reconfigured according to the second selected compensation for loss and damage filter coefficient value set;And
The compensation for loss and damage wave filter that use reconfigures is to process the described Part II of described decoded video frame, with further Improve described decoded video frame and reproduce the informativeness of corresponding uncoded source video frame.
40. methods as claimed in claim 37, wherein, described encoded bit stream also includes more new data and updates instruction, The described instruction described Video Decoder of guidance that updates updates described code book according to described more new data.
41. methods as claimed in claim 40, also include: before configuring described compensation for loss and damage wave filter, by described renewal The compensation for loss and damage filter coefficient value set that market demand is selected in described code book.
42. methods as claimed in claim 41, also include: be decoded period, ringing to the second encoded frame of video follow-up The renewal for described code book that the replacement that Ying Yu is included in described encoded bit stream is applied before indicating and abandoning;
The code book including multiple compensation for loss and damage filter coefficient value set, the plurality of damage is obtained by described video decoder apparatus Evil compensating filter sets of coefficient values is suitable for configuring compensation for loss and damage wave filter with to the decoded video with multifrequency nature Frame processes;
Obtaining encoded bit stream, described encoded bit stream includes described encoded frame of video and selects designator, described choosing Selecting designator makes described Video Decoder can select compensation for loss and damage filter coefficient value set from described code book;
Described encoded frame of video is being decoded period, and described video decoder apparatus selects by described from described code book Select the compensation for loss and damage filter coefficient value set of designator instruction.
43. methods as claimed in claim 37, also include:
Obtain the second code book including more than second compensation for loss and damage filter coefficient value set;And
Described encoded frame of video is being decoded period, according to the selection designator being included in described encoded bit stream Select in described code book and described second code book, select afterwards to be referred to by described selection designator from selected code book The compensation for loss and damage filter coefficient value set shown.
44. 1 kinds for the device being decoded the compensation for loss and damage filter coefficient value set of encoded frame of video, described dress Put and include:
For being obtained the module of the code book including multiple compensation for loss and damage filter coefficient value set, institute by video decoder apparatus State multiple compensation for loss and damage filter coefficient value set to be suitable for configuring compensation for loss and damage wave filter with to the solution with multifrequency nature The frame of code video processes;
For obtaining the module of encoded bit stream, described encoded bit stream includes described encoded frame of video and selects instruction Symbol, described selection designator makes described video decoder apparatus can select compensation for loss and damage filter coefficient from described code book Value set;
For described encoded frame of video being decoded period, selected from described code book by described video decoder apparatus Select the module of the compensation for loss and damage filter coefficient value set indicated by described selection designator;
For configuring described damage by described video decoder apparatus according to selected compensation for loss and damage filter coefficient value set The module of evil compensating filter;
For using the compensation for loss and damage wave filter configured to process described decoded video by described video decoder apparatus Frame at least some of, to improve the module that described decoded video frame reproduces the informativeness of corresponding uncoded source video frame; And
For handled decoded video frame being supplied to the module of the display being associated with described video decoder apparatus.
45. 1 kinds of video decoders, including:
Entropy decoder, is used for obtaining encoded bit stream, and described encoded bit stream includes encoded frame of video and selects instruction Symbol, described selection designator makes described video decoder can select compensation for loss and damage filter coefficient value collection from code book Closing, wherein, described code book includes multiple compensation for loss and damage filter coefficient value set, the plurality of compensation for loss and damage filter coefficient value Set is suitable for configuring compensation for loss and damage wave filter and processes with the frame to the decoded video with multifrequency nature;
Compensation for loss and damage wave filter code book look-up table and address logic, for described encoded frame of video being decoded period, Select from described code book by the compensation for loss and damage filter coefficient value set of described selection designator instruction;And
Compensation for loss and damage filter processor, mends for configuring described infringement according to selected compensation for loss and damage filter coefficient value set Repaying wave filter, wherein, the compensation for loss and damage wave filter configured is for processing at least some of to change of described decoded video frame Kind described decoded video frame reproduces the informativeness of corresponding uncoded source video frame.
The method that 46. 1 kinds of video decoder apparatus realize, for the compensation for loss and damage filter coefficient value to encoded frame of video Set is decoded, and described method includes:
The code book including multiple compensation for loss and damage filter coefficient value set, the plurality of damage is obtained by described video decoder apparatus Evil compensating filter sets of coefficient values is suitable for configuring compensation for loss and damage wave filter with to the decoded video with multifrequency nature Frame processes;
Obtaining encoded bit stream, described encoded bit stream includes described encoded frame of video, selects designator, encoded Multiple adjusted values and adjustment designator, described selection designator makes described Video Decoder can select from described code book Compensation for loss and damage filter coefficient value set, described adjustment designator instructs described Video Decoder according to described encoded multiple Adjusted value adjusts selected compensation for loss and damage filter coefficient value set;
Described encoded frame of video is being decoded period, and described video decoder apparatus selects by described from described code book Select the compensation for loss and damage filter coefficient value set of designator instruction, and adjust institute according to described encoded multiple adjusted values The compensation for loss and damage filter coefficient value set of choosing;
Configured described according to the compensation for loss and damage filter coefficient value set selected by being adjusted by described video decoder apparatus Compensation for loss and damage wave filter;
Used the compensation for loss and damage wave filter configured to process at least the one of decoded video frame by described video decoder apparatus Part, to improve the informativeness of the described decoded video frame corresponding uncoded source video frame of reproduction;And
Handled decoded video frame is supplied to the display being associated with described video decoder apparatus.
47. 1 kinds for the device being decoded the compensation for loss and damage filter coefficient value set of encoded frame of video, described dress Put and include:
For being obtained the module of the code book including multiple compensation for loss and damage filter coefficient value set, institute by video decoder apparatus State multiple compensation for loss and damage filter coefficient value set to be suitable for configuring compensation for loss and damage wave filter with to the solution with multifrequency nature The frame of code video processes;
For obtaining the module of encoded bit stream, described encoded bit stream includes described encoded frame of video, selects instruction Symbol, encoded multiple adjusted values and adjust designator, described selection designator makes the described video decoder apparatus can Selecting compensation for loss and damage filter coefficient value set from described code book, described adjustment designator instructs described video decoder apparatus Selected compensation for loss and damage filter coefficient value set is adjusted according to described encoded multiple adjusted values;
For described encoded frame of video being decoded period, selected from described code book by described video decoder apparatus Select by the compensation for loss and damage filter coefficient value set of described selection designator instruction, and according to described encoded multiple adjusted values Adjust the module of selected compensation for loss and damage filter coefficient value set;
For being joined according to the compensation for loss and damage filter coefficient value set selected by being adjusted by described video decoder apparatus Put the module of described compensation for loss and damage wave filter;
For using the compensation for loss and damage wave filter configured to process decoded video frame by described video decoder apparatus At least partially, to improve the module that described decoded video frame reproduces the informativeness of corresponding uncoded source video frame;And
For handled decoded video frame being supplied to the module of the display being associated with described video decoder apparatus.
48. 1 kinds of video decoders, including entropy decoder, compensation for loss and damage wave filter code book look-up table and address logic, addition Device and compensation for loss and damage filter processor, wherein:
Described entropy decoder is used for obtaining encoded bit stream, and described encoded bit stream includes encoded frame of video, selects to refer to Showing symbol, encoded multiple adjusted values and adjust designator, described selection designator makes the described video decoder can Selecting compensation for loss and damage filter coefficient value set from code book, described adjustment designator instructs described video decoder according to institute Stating encoded multiple adjusted values to adjust selected compensation for loss and damage filter coefficient value set, wherein, described code book includes many Individual compensation for loss and damage filter coefficient value set, the plurality of compensation for loss and damage filter coefficient value set is suitable for configuring compensation for loss and damage Wave filter is to process the decoded video frame with multifrequency nature;
Use described encoded frame of video being decoded period, described compensation for loss and damage wave filter code book look-up table and address logic In selecting from described code book by the compensation for loss and damage filter coefficient value set of described selection designator instruction, and described addition Device is for adjusting selected compensation for loss and damage filter coefficient value set according to described encoded multiple adjusted values;
Described compensation for loss and damage filter processor is for joining according to the compensation for loss and damage filter coefficient value set selected by being adjusted Putting described compensation for loss and damage wave filter, wherein, the compensation for loss and damage wave filter configured is at least to decoded video frame Divide the informativeness carrying out processing to improve the described decoded video frame corresponding uncoded source video frame of reproduction.
CN201180061488.3A 2010-12-21 2011-12-21 Content-adaptive compensation for loss and damage for efficient video coding filters Active CN103339952B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201610854915.8A CN107071484B (en) 2010-12-21 2011-12-21 Device, method and device for decoding video and machine readable medium

Applications Claiming Priority (3)

Application Number Priority Date Filing Date Title
US12/974,645 2010-12-21
US12/974,645 US9462280B2 (en) 2010-12-21 2010-12-21 Content adaptive quality restoration filtering for high efficiency video coding
PCT/US2011/066670 WO2012088359A2 (en) 2010-12-21 2011-12-21 Content adaptive impairments compensation filtering for high efficiency video coding

Related Child Applications (1)

Application Number Title Priority Date Filing Date
CN201610854915.8A Division CN107071484B (en) 2010-12-21 2011-12-21 Device, method and device for decoding video and machine readable medium

Publications (2)

Publication Number Publication Date
CN103339952A CN103339952A (en) 2013-10-02
CN103339952B true CN103339952B (en) 2016-11-30

Family

ID=

Citations (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US5802481A (en) * 1997-03-20 1998-09-01 Motorola, Inc. Adaptive filtering for use with data compression and signal reconstruction

Patent Citations (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US5802481A (en) * 1997-03-20 1998-09-01 Motorola, Inc. Adaptive filtering for use with data compression and signal reconstruction

Similar Documents

Publication Publication Date Title
CN107071484A (en) Content-adaptive compensation for loss and damage for efficient video coding is filtered
CN103430545B (en) Context adaptive motion compensated filtering for efficient video coding
CN106658007B (en) Code and decode method, the coding and decoding equipment of image
KR100957754B1 (en) Image encoding device, image decoding device, image encoding method, image decoding method, computer readable recording medium having image encoding program recorded therein, and computer readable recording medium having image decoding program recorded therein
CN104539969B (en) Image processing apparatus and image processing method
CN103152579B (en) In-loop adaptive wiener filter for video coding and decoding
CN109076225A (en) System and method for transform coefficients encoding
TW200833123A (en) Method and apparatus for encoding image
CN105917650A (en) Block vector prediction in video and image coding/decoding
JP5805097B2 (en) Method and apparatus for DC intra prediction mode for video encoding and decoding
CN102804779A (en) Image processing device and method
KR20040068535A (en) Image encoding apparatus, image decoding apparatus, image encoding method, image decoding method, image encoding program, and image decoding program
CN1835585A (en) Method and system for fast implementation of subpixel interpolation
CN106797483A (en) Use the improvement coded treatment of pallet mode
JP7480303B2 (en) Method and apparatus for video encoding and computer program product thereof
KR20220088503A (en) Inter-picture prediction method and apparatus using a virtual reference image for video coding
JP7467675B2 (en) METHOD AND APPARATUS FOR VIDEO CODING - Patent application
CN102396231A (en) Image-processing device and method
CN103339952B (en) Content-adaptive compensation for loss and damage for efficient video coding filters
CN105007486B (en) Code and decode method, the coding and decoding equipment of image
CN2796278Y (en) DVD decoder
KR20230003566A (en) Block-by-block content-adaptive online training in neural image compression by post-filtering

Legal Events

Date Code Title Description
PB01 Publication
SE01 Entry into force of request for substantive examination
GR01 Patent grant