CN106254870A - Video coding method and system using adaptive color conversion - Google Patents

Abstract

A video coding method and system, the method includes the following steps. An original video frame is received. The original video picture is divided into a coding tree unit (coding tree unit). A coding unit (coding unit) is determined from the coding tree unit. A coding mode (coding mode) of the coding unit is enabled or disabled. If the coding mode is enabled, it is determined whether to estimate a transform unit size in the enabled coding mode. The conversion unit of the coding unit is decided in the enabled coding mode. The size of the coding unit is NxN.

Description

Use method for video coding and the system of self adaptive color conversion

Technical field

It relates to Video coding and coding/decoding method and system.

Background technology

Demand for high quality graphic is gradually increased.Along with the arriving of the video specification such as 4K and 8K, pole need to promote video Coding and decoding efficiency.Additionally, consumer expects to be transmitted by various transmission mediums and receive high quality graphic.Citing For, consumer want to by network in portable device (such as smart mobile phone, tablet PC, mobile computer) and High quality graphic is watched on domestic TV and computer.Consumer also is intended to can be during video conference and Screen sharing Display high quality graphic.

H.265, high efficiency video encoding standard (High Efficiency Video Coding, HEVC) is improving video The coding of compression provides a new standard in decoding usefulness.Compared to original AVC (Advanced Video Coding) standard, by ISO/IEC JTC 1/SC 29/WG 11MPEG (Moving Picture Experts Group) and The HEVC that ITU-T SG16VCEG (Video Coding Experts Group) is set up can reduce compression high-quality video Data transfer rate.AVC standard is the most H.264.

HEVC utilizes inter-frame prediction techniques (Inter prediction) and infra-prediction techniques (Intra Etc. prediction) various coding toolses compress video.Inter-frame prediction techniques utilize video streaming different video picture it Between time redundancy (temporal redundancies) carry out compressed video data.For example, the volume containing Similar content Code and decoded video pictures can be used to encode current video pictures.These encoded and decoded video pictures are permissible It is used for predicting the coding region of current video pictures.Relatively, infra-prediction techniques is merely with in current encoder video pictures Portion's data carry out compressed video data.Infra-prediction techniques does not use the time redundancy of different video picture.For example, at present Video pictures is that the another part utilizing same picture encodes.Infra-prediction techniques includes 35 kinds of frame modes, comprises Plane mode (Planar mode), DC pattern and 33 kinds of directional patterns (directional modes).

Compared to AVC standard, HEVC standard uses expansion cutting techniques (expansive to each input video picture partitioning and dividing).AVC standard when encoding and decoding merely with the big block of input video picture (macroblock) split.On the contrary, HEVC standard can split input video picture becomes various sizes of data cell And block, related description as after.Compared to AVC standard, HEVC standard is to details dynamic, many and the volume of the video pictures of multiple edge Code and decoding program provide more elasticity.

Some coding toolses that can improve video coding program are also listed in HEVC standard.These a little coding toolses are claimed For coding extension (coding extensions).Screen content coding extension (Screen Content Coding Extension, SCC extension) it is absorbed in the treatment efficiency of the video screen content improved under HEVC standard.Screen content The video scene extracted by pattern, word or the video of animation imaging (render) rather than photographing unit.The figure of imaging Case, word or animation can be dynamic or static, and can be provided in the video in the video scene that photographing unit is extracted.SCC Application example can comprise screen mirror (Screen mirroring), high in the clouds game (cloud gaming), Wireless Display Display (displays generated during the access of content (wireless display of content), far-end computer During remote computer desktop access) and Screen sharing (screen sharing) (e.g. video council The instant Screen sharing of view).

A coding tools in SCC is self adaptive color conversion (adaptive color transform, ACT).ACT is It is applied to the color space of the residual pixel sample (residue pixel samples) of coding unit (coding unit, CU) Conversion.For specific color space, there is the color element (color of a pixel of a coding unit (CU) Components) dependency.When the dependency height of the color element of pixel, pixel performs ACT can pass through decorrelation (de-correlating) color element being correlated with is helped to concentrate energy.The practice of this concentration energy can improve coding effect Rate, and reduce coding cost.Therefore, ACT can promote coding usefulness in HEVC cataloged procedure.

But, in an encoding process, need extra code rate distortion function (rate distortion Optimization, RDO) assess whether to enable ACT.RDO is used for assessing code rate distortion (rate distortion, RD) Cost.These evaluation processes may increase encoder complexity and scramble time.Furthermore, when the color element of pixel has gone phase Guan Shi, ACT may the most not necessarily.In this case, owing to performing the cost benefit higher than coding of ACT, further The decorrelation program carrying out color element possibly cannot bring any benefit.

Summary of the invention

One side according to the disclosure, it is provided that a kind of method for video coding.Method for video coding comprises the following steps.Receive One original video frame (source video frame).Segmentation original video frame is a code tree unit (coding tree unit).A coding unit (coding unit) is determined from code tree unit.Enable or disable a coding mode of coding unit (coding mode).If enabling coding mode, then judge whether to estimate a converting unit enabling this coding mode The size of (transform unit).The converting unit of coding unit is determined at the coding mode enabled.The size of coding unit For NxN.

According to another aspect of the present disclosure, it is provided that a kind of video coding system.Video coding system include a memorizer and One processor.Memorizer is in order to store one group of instruction.Processor is in order to perform this group instruction.This group instruction comprises the following steps. Receive an original video frame (source video frame).Segmentation original video frame is a code tree unit (coding tree unit).A coding unit (coding unit) is determined from code tree unit.Enable or disable a coding of coding unit Pattern (coding mode).If enabling coding mode, then judge whether to estimate a converting unit at the coding mode enabled The size of (transform unit).The converting unit of coding unit is determined at the coding mode enabled.The size of coding unit For NxN.

According to another aspect of the present disclosure, it is provided that a kind of non-Transient calculation machine readable recorded medium.Non-Transient calculation machine Readable recorded medium is in order to store one group of instruction.This group instruction is performed by one or more processor, to perform a Video coding Method.This method for video coding comprises the following steps.Receive an original video frame (source video frame).Split former Beginning video pictures is a code tree unit (coding tree unit).A coding unit (coding is determined from code tree unit unit).Enable or disable a coding mode (coding mode) of coding unit.If enabling coding mode, then in the volume enabled Pattern judges whether to estimate the size of a converting unit (transform unit).Determine that coding is single at the coding mode enabled The converting unit of unit.The size of coding unit is NxN.

More preferably understand in order to above-mentioned and other aspect of this disclosure has, multiple embodiments cited below particularly, and coordinate attached Figure, is described in detail below:

Accompanying drawing explanation

Figure 1A～1J illustrates video pictures and the relevant segmentation according to the several embodiment of the disclosure.

Fig. 2 illustrates the video encoder of the disclosure.

Fig. 3 explanation is according to the coded method of the disclosure one embodiment.

Fig. 4 explanation is according to the coded method of another embodiment of the disclosure.

Fig. 5 explanation is according to the coded method of another embodiment of the disclosure.

Fig. 6 explanation is according to the coded method of another embodiment of the disclosure.

Fig. 7 illustrates the calculation flow process of the IPM of non-444 chroma formats.

Fig. 8 illustrates the system of coding and the coding/decoding method performing the disclosure.

[symbol description]

101: video pictures (video frame)

102: code tree unit (coding tree unit, CTU)

103: luminance coding tree block (luma coding tree block, luma CTB)

104:Cb CTB

105:Cr CTB

106,111: related description

107-1,107-2,107-3,107-4: luminance coding block (luma coding block, luma CB)

108: coding unit (Coding unit, CU)

109:Cb CB

110:Cr CB

112: luma prediction block (luma prediction block, PB)

113-1,113-2,113-3,113-4: conversion block (transform block, TB)

114: converting unit (Transform unit, TU)

200: video encoder

202: picture segmentation module (Frame Dividing Module)

204: inter prediction enables ACT module (Inter Prediction enabling adaptive color transformation Module)

206: inter prediction disabling ACT module (Inter Prediction disabling ACT Module)

208: picture depositor (Frame Buffer)

210: mode decision module (Mode Decision Module)

212: infra-frame prediction enables ACT module (Intra Prediction enabling ACT Module)

214: infra-frame prediction disabling ACT module (Intra Prediction disabling ACT Module)

216,218: add up module (Summing Module)

220: switch

222: self adaptive color conversion (ACT) module

224:CCP, conversion and quantization modules (CCP, Transform, and Quantization Module)

226: entropy code module (Entropy Coding Module)

228: inverse operation CCP, conversion and quantization modules (Inverse CCP, Transform, and Quantization Module)

230: switch

232: inverse operation ACT module (Inverse ACT Module)

300,400,500,600,700,800: coded method

304: component correlations analysis (component correlation analysis)

306: outline pattern determines (Rough mode decision)

308: terminate

310: the decision of code rate distortion function mode (rate distortion optimization mode decision, RDO mode decision)

311: whether chroma format is the judgement of non-444 (non-444)

Whether 312:CU size is more than the judgement of marginal value T1

314:TU size determines (TU size ecision)

316: chroma mode determines (chroma mode decision)

Whether 402:CU size is less than the judgement of marginal value T2

702: non-Transient calculation machine readable media

704: processor

Detailed description of the invention

Hereinafter collocation accompanying drawing is explained exemplary embodiment.In the accompanying drawing being described below, unless otherwise saying Bright, the identical label at different accompanying drawings represents the identical or element of approximation.Embodiments presented below not represents the disclosure All performances.It is true that these embodiments are only some examples of the System and method for corresponding to claims.

Figure 1A～1J explanation is according to the video pictures that embodiment of the disclosure and relevant segmentation thereof.

Figure 1A illustrates video pictures 101.Video pictures 101 includes several pixel.Video pictures 101 is divided into several volume Code tree unit (coding tree units, CTUs) 102.The size of each CTU 102 is according to L vertical sample and L water Flat sample (LxL) determines.Each sample corresponds to a pixel value in the different pixels position of CTU.For example, L can be 16,32 or 64.Location of pixels can be pixel position between the position or pixel at CTU place.When location of pixels is pixel Between position, pixel value can be the interpolated value of one or more pixel near location of pixels.Each CTU 102 includes brightness Code tree block (luma coding tree block, luma CTB), chroma coder tree block (chroma CTB) and relevant (associated syntax) is described.

Figure 1B illustrates several CTB and can be contained in a CTU 102 of Figure 1A.For example, CTU 102 can wrap Containing brightness CTB (luma CTB) 103, chrominance C TB (chroma CTB) (containing Cb CTB 104Cr CTB 105).CTU 102 is also Related description (associated syntax) 106 can be included.Cb CTB 104 is blue color difference CTB (blue Difference chroma component CTB), it represents that CTB is in blue change.Cr CTB 105 is red color CTB (red difference chroma component CTB), it represents that CTB is in red change.Related description 106 is wrapped Information and brightness CTB 103, Cb how is coded of containing brightness CTB 103, Cb CTB 104 and Cr CTB 105 The further segmentation of CTB104 and Cr CTB 105.CTB 103, the size of Cb CTB 104 and Cr CTB 105 can be same as The size of CTU 102.Or, the size of brightness CTB 103 can be same as the size of CTU 102, but Cb CTB 104 and Cr The size of CTB 105 can be less than the size of CTU 102.

Infra-frame prediction (intra prediction), inter prediction (inter prediction) and other etc. coding tools Operate in encoded block (coding blocks, CBs).In order to determine that the program of coding is intended to use infra-frame prediction or frame Between predict, CTB can be divided into one or more CB.It is according to four points of tree segmentation (quad-tree that CTB is divided into the program of CB Partitioning) technology.Therefore, CTB can be divided into four CB, and each CB can be divided into four CB again.According to CTB's Size, can proceed such segmentation procedure.

Fig. 1 C illustrate the brightness CTB 103 of Figure 1B be divided into one or more brightness CB 107-1,107-2,107-3 or 107-4.As a example by the brightness CTB of 64x64, corresponding brightness CB 107-1,107-2,107-3 or 107-4 can be the chis of NxN Very little, e.g. 64x64,32x32,16x16 or 8x8.In fig. 1 c, the size of brightness CTB 103 is 64x64.And brightness CTB The size of 103 can be 32x32 or 16x16.

Fig. 1 D illustrates the brightness CTB 103 of Figure 1B and carries out the example that four points of trees are split, and wherein brightness CTB 103 is divided into Brightness CB 107-1,107-2,107-3 or 107-4 of Fig. 1 C.In Fig. 1 D, the size of brightness CTB 103 is 64x64.But, The size of brightness CTB 103 can also be 32x32 or 16x16.

In Fig. 1 D, brightness CTB 103 is divided into the brightness CB 107-2 of four 32x32.The brightness CB of each 32x32 can To be more divided into the brightness CB 107-3 of four 16x16.The brightness CB of each 16x16 can more be divided into the brightness of four 8x8 CB 107-4。

Coding unit (Coding unit, CU) is in order to encode CB.CTB can include only one CU or be divided into number Individual CU.Therefore the size of CU can also be NxN, e.g. 64x64,32x32,16x16 or 8x8.Each CU includes a brightness CB, two chrominance C B and related description.The size of residual error CU produced in encoding and decoding program can be same as its correspondence The size of CU.

Fig. 1 E illustrates the schematic diagram of CB (the brightness CB 107-1 of Fig. 1 C), and these a little CB can be a part of CU 108.Lift For example, CU 108 can include brightness CB 107-1 and chrominance C B (Cb CB 109) and chrominance C B (Cr CB 110).CU 108 can include related description 111.Related description 111 includes how brightness CB 107-1, Cb CB 109 and Cr CB 110 Carry out the information encoded, explanation (brightness CB and the size of chrominance C B, position and further point of e.g. four points tree information Cut).Each CU 108 can have relevant prediction block in brightness CB 107-1, Cb CB 109 and Cr CB 110 (prediction blocks, PBs).Prediction block is combined into predicting unit (prediction units, PUs).

Fig. 1 F illustrate the CB 107-1 of Fig. 1 D be divided into the various of brightness PB 112 may situations.Brightness CB 107-1 is such as It is that the predictability of zones of different according to brightness CB 107-1 is divided into brightness PB 112.For example, brightness CB 107- 1 can comprise single brightness PB 112, and it is equivalently-sized in brightness CB 107-1.Or, brightness CB 107-1 can be vertical Ground or be flatly divided into two even number brightness PB 112.Or brightness CB 107-1 can vertically or horizontally be divided into four Individual brightness PB 112.It should be noted that Fig. 1 F is merely exemplary.The mode of any PB of being divided under HEVC standard all belongs to In the scope of the present disclosure.The depicted mode that brightness CB 107-1 is divided into brightness PB 112 of Fig. 1 F is mutual exclusion.Citing comes Saying, in the intra prediction mode of HEVC, the CB of 64x64,32x32 and 16x16 may be divided into single PB, its size phase It is same as CB.But, the CB of 8x8 may be divided into the PB or the PB of four 4x4 of single 8x8.

Once using infra-frame prediction or inter prediction, produce in different places between prediction block from source video image block Raw residual signals (residual signal) is converted to another territory (domain), to carry out discrete cosine transform further (discrete cosine transform, DCT) or discrete sine transform (discrete sine transform, DST) Coding.In order to provide these to change, each CU or each CB need to utilize one or more conversion block (transform block, TB)。

Fig. 1 G illustrates how the brightness CB 107-1 of Fig. 1 E or Fig. 1 F is divided into different TB 113-1,113-2,113- 3 and 113-4.If brightness CB 107-1 is the CB of 64x64, TB 113-1 be the TB of 32x32, TB 113-2 be the TB of 16x16, TB 113-3 is the TB of 8x8, and TB 113-4 is the TB of 4x4.Brightness CB 107-1 can be divided into 4 TB 113-1,16 TB 113-2,64 TB 113-3 and 256 TB 113-4.One brightness CB 107-1 can be divided into the TB of same size 113 or various sizes of TB 113.

CB is divided into the program of TB according to four points of trees segmentation (quad-tree splitting).Therefore, a CB is permissible Being divided into one or more TB, wherein each TB can further be divided into 4 TB.Such segmentation procedure can basis The size of CB proceeds.

Fig. 1 H illustrates four points of tree segmentations of the brightness CB 107-1 of Fig. 1 E or Fig. 1 F, and it utilizes various partitioning scheme to be divided into TB 113-1,113-2,113-3 or 113-4 of Fig. 1 G.In Fig. 1 H, the size of brightness CB 107-1 is 64x64.But, bright The size of degree CB 107-1 can also be 32x32 or 16x16.

In Fig. 1 H, brightness CB 107-1 is divided into the TB 113-1 of four 32x32.The TB of each 32x32 can be more It is further split into the TB 113-2 of 4 16x16.The TB of each 16x16 can further be divided into the TB of 4 8x8 113-3.The TB of each 8x8 can further be divided into the TB 113-4 of 4 4x4.

TB 113 is then to carry out the conversion of DCT or any HEVC standard.Converting unit (Transform units, TUs) Collect TB 113.One or more TB is used by each CB.CB forms each CU.Therefore, the structure of converting unit (TU) is in difference CU 108 be different, and determined by CU 108.

Fig. 1 I illustrates TB 113-1,113-2,113-3 and 113-4 of the various different segmentations of TU 114.Each TU summary view The TB of 1G or Fig. 1 H segmentation.The TU 114 of 32x32 can use single the TB 113-1's or one or more 16x16 of 32x32 The TB 113-4 of TB 113-3 or 4x4 of TB 113-2,8x8.For using the CU of inter prediction of HEVC, TU can be big In PU so that TU can comprise PU border (boundaries).But, for using the CU of infra-frame prediction of HEVC, TU can Not pass through (cross) PU border.

Fig. 1 J illustrates four points of tree segmentations of the TU 114 of Fig. 1 I, and it utilizes various TB 113-1,113-2,113-3 of Fig. 1 I Or 113-4.In Fig. 1 J, the size of TU 114 is 32x32.But, the size of TU can be 16x16,8x8 or 4x4.

In Fig. 1 J, TU 114 is divided into the TB 113-2 of the TB 113-1 and 4 16x16 of a 32X32.Each The TB of 16x16 can further be divided into the TB 113-3 of 4 8x8.The TB of each 8x8 can further be divided into 4 The TB 113-4 of 4x4.

CTU, CTB, CB, CU, PU, PB, TU or TB described in the disclosure all can include any feature of HEVC standard (feature), size (size) and characteristic (property).Segmentation described in 1C, 1E and 1F figure can also be applied to colourity CTB (Cb CTB 104), chrominance C TB (Cr CTB 105) and chrominance C B (Cb CB 109), chrominance C B (Cr CB 110).

Fig. 2 illustrates the video encoder 200 of the coded method performing the disclosure.Video encoder 200 can include one or Multiple add ons, it provides the additional encoding function of HEVC-SCC, such as color saucer pattern (palette mode), sample certainly Adaptive skew (sample adaptive offset) and block elimination filtering (de-blocking filtering).Additionally, the disclosure Intra prediction mode and the inter-frame forecast mode of other coding modes, e.g. ACT in view of ACT.

Video encoder 200 receives an original video frame (source video frame) of input.This input is original Video pictures first inputs to picture segmentation module (Frame Dividing Module) 202.Picture segmentation module 202 is split former Beginning video pictures is at least one original CT U (source CTU).Original CU (source CU) is obtained by original CT U again.Original The size of CTU and the size of original CU are determined by picture segmentation module 202.Then, encode in the way of by CU.Original After CU is exported by picture segmentation module 202, input to inter prediction enables ACT module (Inter Prediction enabling Adaptive color transformation Module) 204, inter prediction disabling ACT module (Inter Prediction Disabling ACT Module) 206, infra-frame prediction enable ACT module (Intra Prediction enabling ACT Module) 212 and infra-frame prediction disabling ACT module (Intra Prediction disabling ACT Module) 214.

The original CU of input picture is enabled ACT module 204 by inter prediction and encodes, and it utilizes inter-frame prediction techniques and enables Self adaptive color conversion (ACT) determines the prediction of an original CU from input picture.The original CU of input picture is also by inter prediction Disabling ACT module 206 encodes, and it utilizes inter-frame prediction techniques and does not enable self adaptive color conversion (ACT) from input picture certainly The prediction (i.e. disabling ACT) of a fixed original CU.

The reference CU being stored in picture depositor (Frame Buffer) 208 can be used when inter prediction.Original PU And PB is also obtained by original CU, and it is used in inter prediction and enables ACT module 204 and inter prediction disabling ACT module 206 Inter prediction program.Inter prediction utilizes the region of the video pictures of different time to carry out motion detection.Inter prediction enables The encoded inter prediction CU of ACT module 204 and inter prediction disabling ACT module 206 is predefined for high picture quality.Encoded Inter prediction CU is then input to mode decision module (Mode Decision Module) 210.

The original CU of input picture also enables ACT module 212 by infra-frame prediction and encodes, and it utilizes infra-frame prediction skill Art and enable self adaptive color conversion (ACT) and determine the prediction of an original CU from input picture.

The original CU of input picture also encodes by infra-frame prediction disabling ACT module 214, and it utilizes infra-frame prediction skill Art and do not enable self adaptive color conversion (ACT) and determine the prediction (i.e. disabling ACT) of an original CU from input picture.

Infra-frame prediction enables ACT module 212 and infra-frame prediction disabling ACT module 214 when carrying out infra-frame prediction, can use and deposit It is stored in the original CU of the same picture of picture depositor 208.Original PU and PB is also obtained by original CU, and is used in frame pre- Survey and enable ACT module 212 and the process of intra prediction of infra-frame prediction disabling ACT module 214.Encoded infra-frame prediction CU makes a reservation for For high picture quality.ACT module 212 and the encoded frame of infra-frame prediction disabling ACT module 214 output is enabled from infra-frame prediction Interior prediction CU inputs to mode decision module 210.

In mode decision module 210, employing inter prediction enables ACT, inter prediction disables ACT, infra-frame prediction enables The quality of the cost collocation prediction CU that the modes such as ACT and infra-frame prediction disabling ACT carry out the coding of original CU compares.According to Result of the comparison, determines the prediction CU (e.g. inter prediction CU or infra-frame prediction CU) of which coding mode.Selected pre- Survey CU to be then sent to add up module (Summing Module) 216,218.

In adding up module 216, original CU deducts the prediction CU selected, to provide residue CU (residual CU). If the next comfortable inter prediction of the prediction CU selected enables ACT module 204 and infra-frame prediction enables ACT module 121 wherein One of, switch (switch) 220 switches to position A.Input to ACT module (ACT Module) 222 at position A, residue CU, And then input to CCP, conversion and quantization modules 224 (CCP, Transform, and Quantization Module) 224. But, if the prediction CU selected comes comfortable inter prediction disabling ACT module 206 and infra-frame prediction disabling ACT module 214 One of them, switch 220 switches to position B.It is skipped in position B, ACT module 222 and is not carried out in an encoding process. Residue CU is input directly into CCP, conversion and quantization modules 224 from adding up module 216.

In ACT module 222, self adaptive color conversion (adaptive color transform) is implemented in residue CU.ACT The output of module 222 accesses to CCP, conversion and quantization modules 224.

CCP, conversion and quantization modules 224 perform across component prediction (cross component prediction, CCP), Conversion is (such as discrete continuous cosine conversion (Discrete Cosine Transform, DCT) or discrete sine conversion (Discrete Sine Transform, DST) and the quantization of residue CU.The output of CCP, conversion and quantization modules 224 accesses to entropy code mould Block (Entropy Coding Module) 226 and inverse operation CCP, conversion and quantization modules (Inverse CCP, Transform, and Quantization Module)228。

Entropy code module 226 performs Residual Entropy coding (entropy encoding).For example, self adaptation two is entered in full Position arithmetic coding (Context Adaptive Binary Arithmetic Coding, CABAC) can be performed with coding surplus Remaining CU.Any other entropy code program that HEVC is provided all can be implemented in entropy code module 226.

After performing entropy code, the encoded bit stream of the CU of input video picture exports from video encoder 200.Defeated The encoded bit stream gone out can be stored in a memorizer, by transmission line broadcastion or network or provide to a display etc..

In inverse operation CCP, conversion and quantization modules 228, perform being decided by the contrary of CCP, conversion and quantization modules 224 Residue CU, to provide residue CU of a reconstruction.

If the next comfortable inter prediction of the prediction CU selected enables ACT module 204 or infra-frame prediction enables ACT module 212, switch (switch) 230 switches to position C.At position C, residue CU of reconstruction inputs to inverse operation ACT module (Inverse ACT Module) 232 also then inputs to adding up module (Summing Module) 218.But, if selecting Prediction CU come comfortable inter prediction disabling ACT module 206 or infra-frame prediction disabling ACT module 214, switch 230 switches To position D.At position D, inverse operation ACT module 232 is skipped and is not carried out, and residue CU rebuild is directly inputted into and adds Total module 218.

Inverse operation ACT module 232 remains the inverse operation that CU performs the self adaptive color conversion of ACT module 232 to reconstruction. The output input of inverse operation ACT module 232 is to adding up module 218.

In adding up module 218, the prediction CU selected from mode decision module 210 adds residue CU rebuild, To provide the original CU (reconstructed source CU) rebuild.The original CU rebuild is then stored in picture and deposits Device 208, inter prediction and infra-frame prediction for other CU use.

How the coded method 300,400 and 500 of following description is implemented in infra-frame prediction enables in ACT module 212.Coding Method 300,400 and 500 can improve code efficiency and scramble time.

Inter prediction enables ACT module 204, inter prediction disables ACT module 206, infra-frame prediction enables ACT module 212 And infra-frame prediction disabling ACT module 214 is not limited to arrange in a parallel manner.In one embodiment, inter prediction enables ACT module 204, inter prediction disabling ACT module 206, infra-frame prediction enable ACT module 212 and infra-frame prediction disabling ACT module 214 can be with sequential.Inter prediction enables ACT module 204, inter prediction disables ACT module 206, infra-frame prediction enables ACT The arrangement mode of module 212 and infra-frame prediction disabling ACT module 214 can change.

Fig. 3 explanation is according to the coded method 300 of the disclosure one embodiment, and it determines estimation (the TU size of TU size Evaluation) whether need to perform enabling intraframe predictive coding program (the ACT enabled intra prediction of ACT Encoding process) in.More particularly, coded method 300 utilizes the marginal value about CU size to calculate (threshold calculation) decides whether to perform the estimation of TU size.

In step 304, it is former that component correlations analysis (component correlation analysis) is implemented in one Beginning CU, to determine whether the coding mode of the ACT of CU needs to be activated.The dependency of the color component of each pixel in CU is divided Separate out.In each pixel, the dependency of color component and a pixel associated critical value (pixel correlation Threshold) compare, whether be higher than, equal to or less than pixel interdependence marginal value analyzing dependency.

In a CU, calculate the total quantity of pixel higher than pixel interdependence marginal value, wherein equal to pixel interdependence In the pixel of marginal value is also considered as being calculated in higher than pixel interdependence marginal value.The total quantity of pixel is then relevant to a CU Property marginal value (CU correlation threshold) compares.

If the total quantity of pixel is less than CU dependency marginal value, then judge that the color component of CU has low correlation.Therefore, CU is not required to ACT, therefore flow process enters step 308, and at the coding disabling ACT of CU.

But, if the total quantity of pixel is higher than CU dependency marginal value, then judge that the color component of CU has high correlation. In this case, ACT is the component correlations needing each pixel for removing CU.When confirming as high correlation, ACT quilt Enable and flow process enters step 306.In step 306, enable under ACT at infra-frame prediction, carry out an outline pattern and determine.

The correlation analysis of step 304 can be further or optionally according to the color space (color of CU Space) carry out.For example, the color component of the pixel in step 304, CU can be analyzed, and the color of CU is empty Between can judge.Color space can be determined that as redness, green and blue (RGB) space or brightness and colourity (luminance and chrominance, YUV) space.

When being judged to that color space is rgb color space, flow process enters step 306.In step 306, open at infra-frame prediction Under ACT, carry out outline pattern decision (Rough mode decision).Owing to rgb pixel component is generally of high relevant Property, need the dependency of the component carrying out ACT to remove CU each pixel interior, pixel energy (pixel energy) to be isolated For single composition (single component).

Relatively, when color space is judged to YUV color space, and flow process enters step 308, and disables ACT.This be due to YUV pixel component is generally of low correlation, and most pixel energy (pixel energy) is stored in single pixel component (single pixel component).Due to CU pixel component further decorrelation action (de-correlation) no Extra coding efficiency can be produced, therefore be not required to enable ACT in YUV pixel component.

In infra-frame prediction enables ACT module 212, when coded method 300 disables ACT, infra-frame prediction enables the volume of ACT Pattern is disabled, and infra-frame prediction enable ACT module 212 will not export prediction to mode decision module 210.

Enabling ACT module 204 at inter prediction, when inter prediction encoding disabling ACT, inter prediction enables the volume of ACT Pattern is disabled, and inter prediction enables ACT module 204 and will not export prediction to mode decision module 210.

Within step 306, outline pattern decision is carried out under infra-frame prediction enables ACT.Outline pattern determines to be one one-tenth This pattern determines (cost-based mode decision).For example, in outline pattern determines, in that case it can be decided that for low multiple The coding mode of selection of miscellaneous degree cost, quickly to make decision, it is generally of first water and minimum coding cost.

In the step 310, under the coding mode enabling ACT, carry out code rate distortion function mode decision (rate Distortion optimization mode decision, RDO mode decision).Here, when ACT, CCP, conversion, When quantization and entropy code perform, calculate variation (deviation) and the bit cost of coding mode of original video.Make a variation permissible Obtained by erroneous calculations (error calculation), e.g. mean square deviation (mean squared error, MSE).Connect , ROD analyze select the coding mode with minimum coding cost and the highest coding quality by.

For example, in infra-frame prediction enables ACT module 212,35 intra prediction mode (intra prediction Modes, IPMs) it is available for coding.Infra-frame prediction enables ACT module 212 in the outline pattern of step 306 determines, use simple, Low encoding complexity cost determination method goes out minimum coding cost and the highest coding quality person from these Intra prediction mode selections. For example, absolute transformed error and (sum of absolute transform distortion, SATD) cost can be used to Determine the low encoding complexity cost of each IPM.For example, the selection of minimum coding cost and the highest inclined code quality is permissible It is to select 3 IPM or select 8 IPM.Infra-frame prediction enable ACT module 212 step 310 RDO pattern determine in, to respectively Individual IPM has been selected to carry out RDO pattern decision.When ACT, CCP, change, quantify and entropy code perform time, calculate each and select The variation of the original video of IPM and the bit cost of coding.Variation can be obtained by erroneous calculations (error calculation) , e.g. mean square deviation (mean squared error, MSE).Then, analyze from selecting selection IPM to provide by ROD There are minimum coding cost and the IPM of the highest coding quality.

The above-mentioned infra-frame prediction that is relevant to enables the modified flow of ACT module 212 and can also perform to enable ACT at inter prediction Module 204.For example, perform coded method 300 when inter prediction enables ACT module 204, in step 306, carry out time phase The outline pattern of the optimal inter prediction of adjacent video pictures determines, it provides minimum coding cost and the highest coding quality.? Step 310, the RDO pattern carrying out inter prediction determines.Here, when ACT, CCP, change, quantify and entropy code perform time, calculate The variation (deviation) of the original video of inter prediction and coded-bit cost.Variation can be by erroneous calculations (error Calculation) obtain, e.g. mean square deviation (mean squared error, MSE).Then, ROD analysis selection is provided There are minimum coding cost and the inter prediction of the highest coding quality.

In step 312, calculate the CU size of the CU processed at present.The size of CU can be NxN, wherein N can be 4,8, 16,32 or 64.The N value of CU compares with marginal value T1.Marginal value T1 can be 4,8,16,32 or 64.According to comparative result, it is determined that Go out CU size and whether be less than marginal value T1, and thereby estimate the size of the converting unit of coding mode to be enabled.If CU size Less than marginal value T1, flow process enters step 314, to carry out the decision (TU size ecision) of TU size.But, if CU chi Very little equal to or more than marginal value T1, flow process enters step 316, and skips the TU size deciding step of step 314.In step 312, When CU size is more than marginal value T1, determine TU.If CU size CU equals to or more than marginal value T1, TU tetra-sub-tree structure (quadtree structure) may decide that the TU size into maximum possible.For example, when CU size is equal to or more than facing Dividing value T1, for the PU of a size of 64x64, in that case it can be decided that go out the TU of four 32x32.In another embodiment, when CU size etc. In or more than marginal value T1, for the PU of 32x32,16x16,8x8 or 4x4, TU can be with PU same size.Citing comes Saying, if the size of PU is 32x32, corresponding PU size can be 32x32.

Owing to the decision of TU size expends the time and increases coding cost, step 312 can improve scramble time and efficiency.Cause This, if the decision of TU size can be omitted, then can Xie Sheng coding cost instant between.Furthermore, CU size equals to or more than marginal value T1 represents the content of CU uncomplicated.For example, more than marginal value T1, CU size may represent that video image has district on a large scale Territory does not has border, dynamic or complicated image.Therefore, the decision of TU size can be made without, and regards carrying out height efficiently Frequently the coding of the CU of quality.

In a step 314, if CU size is less than marginal value T1, then the decision of TU size is performed.Here, determine original CU TU.By the RDO cost estimate of step 310, analyze TU size, obtained the CU's of peak efficiency and high video quality ACT changes.For example, the TU size of 4x4,8x8,16x16 and 32x32 can be analyzed.As the ACT being obtained in that peak efficiency The TU size of conversion is determined out, and this TU size is selected to make the ACT of CU and changes and enter step 316.Select TU size is as optimal TU tetra-sub-tree structure size.

In step 316, carry out chroma mode decision (chroma mode decision).The decision of chroma mode is based on The decision of the predictive mode of step 310 is carried out, and utilizes determined predictive mode (determined prediction Mode) colorimetric prediction (chroma prediction) is made to produce colourity PU (chroma PU) and colourity TU of correspondence (chroma TU).The TU determined from step 312 or step 314 also can be used to produce colourity TU.Colourity TU is also according to chroma format (chroma format) carries out double sampling (subsample).Therefore, in one embodiment, when chroma format is 4:2:0, and The size of brightness TU is 32x32, and colourity TU of decision is colourity TU of 16x16.

In step 308, infra-frame prediction enables the selection optimum frame inner estimation mode of ACT module and selects optimal TU tetra-points tree The program of physical dimension completes.Prediction and RDO cost produce, and input is to mode decision module 210, to predict with other Module input compares to the RDO cost of mode decision module 210.For example, inter prediction enable ACT module 204 can Prediction and the RDO cost of the CU enabling ACT can be produced, and input prediction CU and RDO cost are to mode decision module 210.Interframe Prediction disabling ACT module 206 and infra-frame prediction disabling ACT module 214 also produce prediction CU and RDO cost, and input it each Prediction CU and RDO cost to mode decision module 210.Mode decision module 210 compare inter prediction enable ACT module 204, Inter prediction disabling ACT module 206, infra-frame prediction enable ACT module 212 and infra-frame prediction disables what ACT module 214 was inputted Prediction CU and RDO cost, and determine the prediction CU that will input to adding up module 216,218.

Fig. 4 explanation is according to the coded method 400 of another embodiment of the disclosure, and it determines according to another embodiment of the disclosure ACT is the need of enabling.More particularly, coded method 400 make use of the marginal value about CU size to calculate (threshold And the decision of dependency of color component of CU pixel calculation).Calculating according to marginal value, ACT can enable or prohibit With.The element of identical label refers to aforementioned related description.

In step 304, component correlations analysis (component correlation analysis) is implemented in original CU, To determine whether ACT needs to enable or disable.Step 304 is such as the explanation of coded method 300.If the dependency of the color component of CU Height, then enable ACT and flow process enter step 306,310,314,316 and 308 (with above-mentioned coding step 300).But, if relevant Property low, then flow process enter step 402.

In step 402, determine the size of the CU processed at present.As it was previously stated, CU a size of NxN, wherein N can be 4,8, 16,32 or 64.The N value of CU compares with marginal value T2, whether to compare CU size less than marginal value T2.Marginal value T2 can Being 4,8,16,32 or 64.If CU is smaller in size than marginal value T2, then enables ACT and flow process enters step 310, such as coding staff The RDO pattern of the step 310 of method 300 determines.But, if CU size enters step 308 equal to or more than marginal value T2, flow process, And disable ACT.

ACT module 204 is enabled, when in coded method 400, ACT is disabled, and inter prediction enables ACT module at inter prediction 204 are output as not applying the inter prediction CU of ACT.Therefore, in this case, inter prediction enables ACT module 204 and exports CU be same as inter prediction disabling ACT module 206 output.Similarly, enable ACT module 212 at infra-frame prediction, work as coding In method 400, ACT is disabled, and infra-frame prediction enables ACT module 212 and is output as not applying the infra-frame prediction CU of ACT.Therefore, In this case, infra-frame prediction enables the output CU of ACT module 212 and is same as the defeated of infra-frame prediction disabling ACT module 214 Go out.

Owing to CU is equivalently-sized or represents that the content of CU is uncomplicated more than marginal value T2, step 402 can be improved the scramble time And code efficiency.More than marginal value T2, CU size may represent that video image has extensive area not have border, dynamic or complicated Image.Under the color component of combination fully decorrelation, in order to encode CU efficiently, it may not be necessary to ACT.

Fig. 5 explanation is according to the coded method 500 of another embodiment of the disclosure, and it determines according to another embodiment of the disclosure ACT is the need of enabling and carrying out TU size estimation the need of being calculated by two marginal values.More particularly, coding Method 500 uses the first marginal value about CU size to calculate (first threshold calculation) and in order to sentence The dependency of the CU color components in pixels of disconnected ACT to be enabled determines.Coded method 500 also uses second about CU size Marginal value calculates (second threshold calculation), to determine whether the estimation of TU size needs to perform.Identical mark Number element refer to aforementioned related description.

In step 304, component correlations analysis (component correlation analysis) is implemented in original CU, To determine whether ACT needs to enable or disable.Step 304 is such as the explanation of coded method 300.If the dependency of the color component of CU Height, then enable ACT and flow process enters step 306, to carry out the RDO pattern decision of outline pattern decision and step 310.Step 306 and 310 such as the narration of aforementioned encoding methods 300.But, if dependency is low, then flow process enters step 402.

In step 402, determine the size (as described in the coded method 400 of earlier figures 4) of the CU processed at present.If CU size Less than marginal value T2, then enable ACT, and enter step 310, to carry out RDO pattern decision.But, if CU size is equal to or big In marginal value T2, then flow process enters step 308, and disables ACT.

ACT module 204 is enabled, when in coded method 500, ACT is disabled, and inter prediction enables ACT module at inter prediction 204 are output as not applying the inter prediction CU of ACT.Therefore, in this case, inter prediction enables ACT module 204 and exports CU be same as inter prediction disabling ACT module 206 output.

Similarly, enable ACT module 212 at infra-frame prediction, when in coded method 500, ACT is disabled, and infra-frame prediction enables ACT module 212 is output as not applying the infra-frame prediction CU of ACT.Therefore, in this case, infra-frame prediction enables ACT module The output CU of 212 is same as the output of infra-frame prediction disabling ACT module 214.

In step 310, RDO pattern determines the content as described in aforementioned encoding methods 300.

In step 312, the calculating of the CU size processed at present content as described in aforementioned encoding methods 300, determine Whether CU size is less than marginal value T1.If CU is smaller in size than marginal value T1, then flow process enters step 314, to carry out TU size certainly Fixed.But, if CU size is equal to or more than marginal value T1, flow process enters step 316, and the TU size skipping step 314 determines. Step 314,316 decision process such as aforesaid coded method 300.

Marginal value T1 and T2 can be set as identical or different value.

The coded method 500 of Fig. 5 combines marginal value calculating and improves code efficiency and time.As it has been described above, CU size etc. In or represent that the content of CU is uncomplicated more than marginal value T2, and the non-boundary, dynamically or complicated figure of extensive area can be contemplated that Sample.Under the color component of combination fully decorrelation, in order to encode CU efficiently, it may not be necessary to ACT.Furthermore, After the TU size of step 314 determines to be omitted, it is possible to save coding cost.

Fig. 6 explanation is according to the coded method 600 (being similar to coded method 300) of another embodiment of the disclosure, and it is according to this Another embodiment disclosed decides whether to perform TU size estimation in the process of intra prediction enable ACT.More specifically come Saying, weave method 600 uses the marginal value about CU size to calculate (threshold calculation), and according to marginal value Calculating judges whether to need to perform TU size estimation.

In step 304, component correlations analysis (component correlation analysis) is implemented in original CU, To determine whether ACT needs to enable or disable.Step 304 is such as the explanation of coded method 300.If the dependency of the color component of CU Height, then enable ACT and flow process enters step 306, to carry out the RDO pattern decision of outline pattern decision and step 310.Step 306 and 310 such as the narration of aforementioned encoding methods 300.But, if the dependency in step 304 is low, or color space is judged as YUV color space, then enable the coding mode of ACT and be directly entered step 310, but does not perform the outline pattern of step 306 certainly Fixed.Here, for low correlation pixel component or YUV color space, ACT still enables, to confirm the decorrelation of pixel component Property may produce additional coding efficiency.

In step 310, the calculating that RDO pattern determines is such as aforementioned encoding methods 300.

In step 312, the calculating of the CU size processed at present content as described in aforementioned encoding methods 300, determine Whether CU size is less than marginal value T1.If CU is smaller in size than marginal value T1, then flow process enters step 314, to carry out TU size certainly Fixed.But, if CU size is equal to or more than marginal value T1, flow process enters step 316, and the TU size skipping step 314 determines. Step 314,316 decision process such as aforesaid coded method 300.

Marginal value T1 and T2 can be set as identical or different value.

The coding/decoding method of the reverse procedure performing coded method 300,400,500,600 can be efficiently to coded method 300, the video coded by 400,500,600 is decoded.Therefore, the foregoing of the disclosure be enough to understand execution coded method 300, the coding/decoding method of the reverse procedure of 400,500,600.Disclosure foregoing also be enough to understand to coded method 300, 400, the video coded by 500,600 is decoded other required decoding programs.

If big CU uses IPM as screen virtual content (screen visual content), then may represent this district The content in territory is the most uncomplicated, and is not required to estimate the size of TU.Therefore, the IPM of non-444 chroma formats is prohibited the big CU of part TU segmentation.Fig. 7 illustrates the calculation flow process of the IPM of non-444 chroma formats.Step 306 and 310 is such as aforementioned encoding methods 300 Narration.In step 310, the calculating that RDO pattern determines is such as aforementioned encoding methods 300.

In step 311, it is judged that whether chroma format is non-444.If chroma format is non-444, then enter step 312.If Colourity lattice are not non-444, then enter step 314, determine with nearly row TU size.

In step 312, the calculating of the CU size processed at present content as described in aforementioned encoding methods 300, determine Whether CU size is less than marginal value T1.If CU is smaller in size than marginal value T1, then flow process enters step 314, to carry out TU size certainly Fixed.But, if CU size is equal to or more than marginal value T1, flow process enters step 316, and the TU size skipping step 314 determines. Step 314,316 decision process such as aforesaid coded method 300.

Marginal value T1 and T2 can be set as identical or different value.

Fig. 8 illustrates the system 700 of coding and the coding/decoding method performing the disclosure.System 700 includes a non-Transient calculation machine Readable media (non-transitory computer-readable medium) 702, it can be storage array instruction Memorizer.This instructs a bit and can be performed by processor 704.It should be noted that one or more non-Transient calculation machine readable media 702 and/or one or more processor 704 can optionally use, to perform coding and the coding/decoding method of the disclosure.

Non-Transient calculation machine readable media 702 can be any kind of non-Transient calculation machine readable recorded medium (non-transitory computer-readable storage medium, non-transitory CRM).Non-transient state meter Calculation machine readable recorded medium can include floppy discs (floppy disk), pliability disc (flexible disk), hard Dish (hard disk), Winchester disk drive (hard drive), solid state hard disc (solid state drive), tape (magnetic Tape), any magnetic data storage medium (magnetic data storage medium), CD-ROM device (CD-ROM), any Optical data carrier (optical data storage medium), any physical property medium, dynamic with pattern of holes State random access memory (RAM), programmable read only memory (PROM), EPROM (EPROM), fast Dodge EPROM (FLASH-EPROM), any flash memory, nonvolatile memory (NVRAM), fast Take (cache), depositor (register), memory chip (memory chip), film (cartridge) and network.Calculate Machine readable recorded medium can store the array instruction performed by least one processor.This instructs a bit to comprise and makes processor go to perform The coding of the disclosure and the step of coding/decoding method or stage.Furthermore, one or more computer-readable recording medium can be used to Realize coding and the coding/decoding method of the disclosure." computer-readable recording medium " comprises visible object but does not comprise carrier carrier wave Signal and transient signal.

Processor 704 can be any type of digital signal processor (digital signal processor, DSP), ASIC (application specific integrated circuit, ASIC), digital signal Processing means (digital signal processing device, DSPD), programmable logic device (programmable Logic device, PLD), programmable logic array (field programmable gate arrays, FPGA), controller (controller), microcontroller (micro-controller), microprocessor (micro-processor), computer or appoint What he is able to carry out the coding electronic component with coding/decoding method of the disclosure.

Experimental result

The experimental result of the coded method of the following description disclosure.

Laboratory herein use HEVC SCC reference model, general test condition (common test conditions, CTC) SCM 4.0 under.The coding usefulness of the coded method of the disclosure is that the reference model with HEVC compares.HEVC reference Model takes scramble time A and encodes.The Test code method of the disclosure takes and carries out scramble time B and carry out Coding.Scramble time percentage ratio is that scramble time B is divided by scramble time A.Experiment can use HEVC general test flow process.Video can Mixing word, image, dynamic menu, mixed content, animation, photographing unit extract content.Video can be have 720p, 1080p, Or the rgb color space of the image quality of 1440p and YUV color space.Experiment employing damages under condition (lossy condition) Full infra-frame prediction, random access memory and low B predict (low-B prediction).Full infra-frame prediction uses the picture currently compressed Information in face compresses video pictures, and random access memory and low B prediction use the most encoded picture and currently pressed The information of the picture of contracting compresses video pictures.In the following description, low B prediction may also mean that low latency B predicts (low delay B prediction).In each experiment, scramble time and decoding time are all got off with percentage registration, this A little percentage ratios represent the ratio of the coded method relative to reference model and coding/decoding method.Relative to original video source, about respectively The positive percentage ratio of individual G/Y, B/U and R/V component represents that bit rate coding loses (bit rate coding loss), negative hundred Proportion by subtraction represents bit rate coding gain (bit rate coding gain).For example, 0.1% numeric representation of G/Y component The G/Y component of encoded video is 0.1% relative to the coding loss of the G/Y component of original video.In another example, G/Y The G/Y component of-0.1% encoded video of numeric representation of component relative to the coding gain of the G/Y component of original video is 0.1%.

Refer to coded method 500 and the table 1 below of Fig. 5.In coded method 500, laboratory is implemented in following three kinds of settings Under.Setting one, marginal value T2 and marginal value T1 are all set as 64.Setting two, marginal value T2 is set as 64, marginal value T1 It is set as 32.Setting three, marginal value T2 is set as 64, and marginal value T1 is set as 16.Infra-frame prediction is predetermined coding mode.

Setting one, when pixel component has low correlation, size is more than or equal to the CU of 64x64 by not enable ACT Mode encode.The CU being smaller in size than 64x64 is encoded in the way of enabling ACT.Furthermore, in the CU size feelings more than 64x64 Under condition, omit the step 314 that TU size determines.In the case of 64x64 is smaller in size than for CU, perform the step that TU size determines 314。

Setting two, when pixel component has low correlation, size is more than or equal to the CU of 64x64 by not enable ACT Mode encode.The CU being smaller in size than 64x64 is encoded in the way of enabling ACT.Furthermore, in the CU size feelings more than 32x32 Under condition, omit the step 314 that TU size determines.In the case of 32x32 is smaller in size than for CU, perform the step that TU size determines 314。

Setting three, when pixel component has low correlation, size is more than or equal to the CU of 64x64 by not enable ACT Mode encode.The CU being smaller in size than 64x64 is encoded in the way of enabling ACT.Furthermore, in the CU size feelings more than 16x16 Under condition, omit the step 314 that TU size determines.In the case of 16x16 is smaller in size than for CU, perform the step that TU size determines 314。

Table 1

As shown in table 1, the coding usefulness setting one, setting two and setting three all has improvement.Set one and reduce the volume of 3% Code complexity (encoding complexity), sets two and reduces the encoder complexity of 6%.Set three and reduce the volume of 9% Code complexity (sets three reductions at most).Therefore, all of setting can improve code efficiency.Each is set in bit rate Under least disadvantage (minimal loss of bit rate), scramble time and efficiency have improvement.

Refer to coded method 500 and table 2 below, 3.Here, experiment is in full frame, random access memory and low latency B (low Delay B) under carry out.In experiment one, marginal value T2 and marginal value T1 are all set as 32.In experiment two, marginal value T2 and facing Dividing value T1 is all set as 16.Such as coded method 500, in experiment one, size is estimated more than or equal to the CU disabling TU of 32x32 (TU evaluation), and size encodes more than or equal to 32x32 in the way of not enabling ACT.In experiment two, size is more than Or estimate equal to the CU disabling TU of 16x16, and size encodes in the way of not enabling ACT more than or equal to 16x16.It is smaller in size than The CU of 16x16 encodes in the mode enabling ACT.Experiment is to damage condition (lossy condition) and full frame frame inner region Carry out under block reproduction technology (full frame intra block copy).

Table 2

As described in Table 2, in experiment one, full frame internal schema (all intra mode) reduces the encoder complexity of 5%.With Machine access and low latency B respectively reduce the encoder complexity of 1%.Each setting demonstrates low-down bit rate penalty, full frame In and random access memory almost without change bit rate.

In experiment two, full frame internal schema reduces the encoder complexity of 8%.The coding that random access memory reduces 1% is complicated Degree.Low latency B does not change encoder complexity.Compared to experiment one, each pattern has a more bit rate penalty, but bit Rate loss remains within minimum (only in the range of the decimal of percentage ratio).Compared to original video, encoded video is the most slightly Reduce bit rate, therefore only lose least a portion of video quality.Owing to coded method 500 improves the scramble time, therefore such regard Frequently quality is acceptable in most application.

Table 3

As described in Table 3, in experiment one and experiment two, each pattern is all or on average, bit rate does not the most change Become.The encoder complexity (all reducing by 1% in each experiment) of most ratio is reduced in full frame.

Refer to coded method 500 and the table 4 below of Fig. 5.Here, experiment be damage condition (lossy condition), Carry out under 4-CTU frame intra block reproduction technology (Intra block copy) and 4:4:4 chroma mode.Frame intra block reproduction technology Motion vector is utilized to replicate a block to current encoder video pictures from previous encoded CU.4-CTU points out that motion vector can The scope searched.

In experiment one, marginal value T2 and marginal value T1 are all set as 32.In experiment two, marginal value T2 and marginal value T1 all set It is set to 16.Such as coded method 500, in experiment one, size is estimated more than or equal to the CU disabling TU of 32x32.At experiment two, chi The very little CU more than or equal to 16x16 disables TU and estimates.In experiment one, size enables ACT more than the CU of 32x32, size more than or ACT is disabled equal to the CU of 32x32.In experiment two, the CU being smaller in size than 16x16 enables ACT, and size is more than or equal to 16x16's CU disables ACT.

Table 4

As described in Table 4, in experiment one and experiment two, in full frame, random access memory or low latency B-mode be all minimal bit rate Change.Reducing most encoder complexities in full frame, it reduces 5% in experiment one, reduces 8% in experiment two.

Refer to the coded method 400 of Fig. 4 and table 5 below .1 and table 5.2.Here, marginal value T2 is set as 64.Cause This, when the color component that the component correlations analysis of step 304 analyzes CU has low correlation, perform step 402, to sentence Whether disconnected CU size is less than 64x64.If CU is smaller in size than 64x64, then enables ACT and perform the RDO pattern decision of step 310. If CU size is more than or equal to 64x64, then disabling ACT and entrance step 308.Experiment one employing full frame frame intra block replicates skill Art (full frame intra block copy) damage full intra-frame encoding mode (lossy all intra encoding Mode), that tests two employing 4CTU IBC technology damages full intra-frame encoding mode.Chroma mode is 4:4 in each experimental selection: 4。

Table 5.1: experiment one

Table 5.2: experiment two

As shown in table 5.1, in YUV color space and full frame, damage, under full frame frame intra block reproduction technology, coding Method 400, under minimal bit rate loses, reduces the scramble time of 1% to 3%.As shown in table 5.2, in full frame, damage, Under 4CTU frame intra block reproduction technology, coded method 400 is under minimum bit is lost, and the ratio reducing the scramble time is similar to table The experiment one of 5.1.

Refer to coded method 400 and table 6 below.Here, marginal value T2 is set as 64.Chroma mode at 4:4:4 performs Lossless intraframe coding (Lossless intra encoding).

Table 6

At YUV color space, coded method saves the scramble time of 0% to 2%.

Refer to coded method 300 and the table 7 below of Fig. 3.Here, marginal value T1 is set as 32 in experiment one, in experiment two It is set as 16.Such as coded method 300, in experiment one, when CU size is more than or equal to 32x32, the TU chi of step 314 will be omitted Very little decision；When CU is smaller in size than 32x32, then the TU size performing step 314 determines.In experiment two, CU size is more than or equal to During 16x16, the TU size omitting step 314 is determined；When CU is smaller in size than 16x16, then perform the TU size of step 314 certainly Fixed.What experiment performed to enable ACT damages full intraframe coding.

Table 7

The scramble time of experiment one saves 3% to 6%.The scramble time of experiment two saves 6% to 10%.Therefore, Only it is less than under 32x32 or 16x16 in CU size and just allows to carry out TU size decision, to help code efficiency.

Foregoing in order to the technology of the disclosure to be described, so itself and be not used to limit to present disclosure.Repairing of embodiment Change and all fall within the scope of the present disclosure with adjusting.For example, the disclosed embodiments comprise software and hardware, but the disclosure is System can only realize with hardware with method.

Software developer can develop a computer program based on disclosed method, and it can use various computer program skill Art is developed.For example, usability of program fragments or program module can be by Java, C, C++, compositional language or any other programs Language is developed.One or more pieces of software is with module can be installed on a computer system, non-Transient calculation machine can read Jie Matter or existing communication software.

Furthermore, although above-mentioned to disclose various embodiment, right the scope of the present disclosure comprise the equalization of various element, amendment, Omit, combine (combinations between such as different embodiments), application or select.The element of claims is with the widest scope Lay down a definition, and be not limited to the content of embodiment.(comprise adjustment order additionally, the step of method can be modified, insert Or delete step).Although the disclosure is open as above with preferred embodiment, so it is not limited to the disclosure.The disclosure Protection domain is when being as the criterion depending on appended claims confining spectrum.

Those skilled in the art of the invention understand other embodiment also dependent on the explanation of the disclosure.The model of the disclosure Enclose comprise combine general knowledge various changes, implement with application.Description is merely exemplary with embodiment, the guarantor of the disclosure Scope of protecting is when being as the criterion depending on appended claims confining spectrum.

Claims (17)

1. a method for video coding, including:

Receive original video frame (source video frame)；

Splitting this original video frame is code tree unit (coding tree unit)；

Coding unit (coding unit) is determined from this code tree unit；

Enable or disable the coding mode (coding mode) of this coding unit；

If enabling this coding mode, then judge whether to estimate converting unit (transform unit) at this coding mode enabled Size；And

This converting unit of this coding unit is determined at this coding mode enabled；

Wherein the size of this coding unit is NxN.

2. the method for claim 1, also includes:

Judge the color space of this coding unit；

Wherein judge that this color space of this coding unit includes judging whether this color space is red, green and blue (RGB) color space or brightness and colourity (YUV) color space.

3. method as claimed in claim 2, also includes:

If this coding mode enabled is for enabling the frame of self adaptive color conversion (adaptive color transform, ACT) Inner estimation mode (intra prediction mode), when this color space is judged as this rgb color space, performs into This pattern determines (cost-based mode decision).

4. method as claimed in claim 2, also includes:

If this coding mode enabled is for enabling the frame of self adaptive color conversion (adaptive color transform, ACT) Inner estimation mode (intra prediction mode), when this color space is judged as brightness and colourity (YUV) color space Time, executory cost pattern determines (cost-based mode decision).

5. method as claimed in claim 2, also includes:

When this color space is judged as this YUV color space and N more than or equal to marginal value, disable this volume of this coding unit Pattern.

6. method as claimed in claim 2, also includes:

Judge that whether N is less than marginal value；And

When this color space is judged as this YUV color space and N less than this marginal value, enable this coding mode, this coding mould Formula enables self adaptive color conversion (adaptive color transform, ACT).

7. method as claimed in claim 2, also includes:

Judge that whether N is less than the first marginal value；

Judge that whether N is less than the second marginal value；

When color space is judged as this YUV color space and N less than this first marginal value, enable this coding mode, this coding Pattern enables self adaptive color conversion (adaptive color transform, ACT)；And

When N is less than this second marginal value, estimating the size of this converting unit at this coding mode, it is adaptive that this coding mode enables this Property color conversion.

8. method as claimed in claim 2, also includes:

Judge that whether N is less than the first marginal value；

Judge that whether N is more than or equal to the second marginal value；

When this color space is judged as this YUV color space and N less than this first marginal value, enable this coding mode, this volume Pattern enables self adaptive color conversion (adaptive color transform, ACT)；And

When N is more than or equal to this second marginal value, not estimating the size of this converting unit at this coding mode, this coding mode opens By this self adaptive color conversion.

9. the method for claim 1, also includes:

Judge that whether N is more than or equal to marginal value；And

When N enables self adaptive color conversion (adaptive color more than or equal to this marginal value and this coding mode of enabling Transform, ACT), do not assess the size of this converting unit.

10. the method for claim 1, also includes:

Judge that whether N is less than marginal value；

When N enables self adaptive color conversion (adaptive color less than this marginal value and this coding mode of enabling Transform, ACT), assess the size of this converting unit；And

Select the size of this converting unit.

11. the method for claim 1, also include:

If this original video frame be non-444 (non-444) and this coding unit be smaller in size than NxN, assess this converting unit Size.

12. 1 kinds of video coding systems, including:

Memorizer, in order to store group instruction；And

Processor, in order to perform the instruction of this group, the instruction of this group includes:

Receive original video frame (source video frame)；

Splitting this original video frame is code tree unit (coding tree unit)；

Coding unit (coding unit) is determined from this code tree unit；

Enable or disable the coding mode (coding mode) of this coding unit；

If enabling this coding mode, then judge whether to estimate converting unit (transform unit) at this coding mode enabled Size；And

This converting unit of this coding unit is determined at this coding mode enabled；

Wherein the size of this coding unit is NxN.

13. systems as claimed in claim 12, wherein this processor also includes in order to this group instruction performed:

Judge the color space of this coding unit；

Wherein this color space is judged if it is redness, green and blueness (RGB) color space or brightness and colourity (YUV) color Color space.

14. systems as claimed in claim 13, wherein this processor also includes in order to this group instruction performed:

Judge that whether N is less than the first marginal value；

Judge that whether N is less than the second marginal value；

When color space is judged as this YUV color space and N less than this first marginal value, enable this coding mode, this coding Pattern enables self adaptive color conversion (adaptive color transform, ACT)；And

When N is less than this second marginal value, estimating the size of this converting unit at this coding mode, it is adaptive that this coding mode enables this Property color conversion.

15. systems as claimed in claim 13, wherein this processor also includes in order to this group instruction processed:

Judge that whether N is less than the first marginal value；

Judge that whether N is more than or equal to the second marginal value；

When this color space is judged as this YUV color space and N less than this first marginal value, enable this coding mode, this volume Pattern enables self adaptive color conversion (adaptive color transform, ACT)；And

When N is more than or equal to this second marginal value, not estimating the size of this converting unit at this coding mode, this coding mode opens By this self adaptive color conversion.

16. systems as claimed in claim 12, wherein this processor also includes in order to this group instruction processed:

If this original video frame be non-444 (non-444) and this coding unit be smaller in size than NxN, assess this converting unit Size.

17. 1 kinds of non-Transient calculation machine readable recorded mediums, in order to store group instruction, this group instructs by one or more processor Performing, to perform method for video coding, wherein this method for video coding includes:

Receive original video frame (source video frame)；

Splitting this original video frame is code tree unit (coding tree unit)；

Coding unit (coding unit) is determined from this code tree unit；

Enable or disable the coding mode (coding mode) of this coding unit；

If enabling this coding mode, then judge whether to estimate converting unit (transform unit) at this coding mode enabled Size；And

This converting unit of this coding unit is determined at this coding mode enabled；

Wherein the size of this coding unit is NxN.