Methods and apparatuses for video coding using triangle partition
Patent Information
- Application Number
- IN202448045097
- Authority / Receiving Office
- IN · IN
- Patent Type
- Patents
- Current Assignee / Owner
- Priority Date
- 2019-10-10
- Filing Date
- 2024-06-11
- Publication Date
- 2026-10-13
- Estimated Expiration
- 2040-10-12
AI Technical Summary
Current video coding techniques face challenges in efficiently encoding and decoding high-definition video data, particularly in maintaining image quality while reducing the exponential growth of video data, especially with the introduction of higher resolutions like 4Kx2K or 8Kx4K, which demands more efficient encoding and decoding methods.
The implementation of geometric partitioning in video coding, where video pictures are divided into coding units (CUs) that can be further partitioned into geometric-shaped prediction units (PUs), allowing for the construction of merge lists using motion vectors from List 0 and List 1, enabling the selection of uni-prediction motion vectors based on binary reference list indication flags to improve prediction accuracy.
This approach enhances coding efficiency and maintains image quality by allowing for more flexible and accurate prediction methods, specifically in high-definition and higher resolution video formats, by utilizing geometric partitions and advanced motion vector selection techniques.
Abstract
Description
FIELD
[0001] The present application generally relates to video coding andcompression, and in particular but not limited to, methods and apparatus for motioncompensated prediction using triangular prediction unit (i.e. a special case ofgeometric partition prediction unit) in video coding.BACKGROUND
[0002] Digital video is supported by a variety of electronic devices, such asdigital televisions, laptop or desktop computers, tablet computers, digital cameras,digital recording devices, digital media players, video gaming consoles, smartphones, video teleconferencing devices, video streaming devices, etc. Theelectronic devices transmit, receive, encode, decode, and / or store digital video databy implementing video compression / decompression. Digital video devicesimplement video coding techniques, such as those described in the standardsdefined by Versatile Video Coding (VVC), Joint Exploration Test Model (JEM),MPEG-2, MPEG-4, ITU-T H.263, ITU-T H.264 / MPEG-4, Part 10, AdvancedVideo Coding (AVC), ITU-T H.265 / High Efficiency Video Coding (HEVC), andextensions of such standards.
[0003] Video coding generally utilizes prediction methods (e.g., inter-prediction,intra-prediction) that take advantage of redundancy present in video images orsequences. An important goal of video coding techniques is to compress video datainto a form that uses a lower bit rate, while avoiding or minimizing degradations tovideo quality. With ever-evolving video services becoming available, encodingtechniques with better coding efficiency are needed. Block partitioning schemes ineach standard are also evolving.
[0004] Video compression typically includes performing spatial (intra frame)prediction and / or temporal (inter frame) prediction to reduce or remove redundancyinherent in the video data. For block-based video coding, a video frame ispartitioned into one or more slices, each slice having multiple video blocks, whichmay also be referred to as coding tree units (CTUs). Each CTU may contain onecoding unit (CU) or recursively split into smaller CUs until the predefinedminimum CU size is reached. Each CU (also named leaf CU) contains one ormultiple transform units (TUs) and each CU also contains one or multipleprediction units (PUs). Each CU may be coded in intra, inter or IBC modes. Videoblocks in an intra coded (I) slice of a video frame are encoded using spatialprediction with respect to reference samples in neighbor blocks within the samevideo frame. Video blocks in an inter coded (P or B) slice of a video frame may usespatial prediction with respect to reference samples in neighbor blocks within thesame video frame or temporal prediction with respect to reference samples in otherprevious and / or future reference video frames.
[0005] Spatial or temporal prediction based on a reference block that has beenpreviously encoded, e.g., a neighbor block, results in a predictive block for a currentvideo block to be coded. The process of finding the reference block may beaccomplished by block matching algorithm. Residual data representing pixeldifferences between the current block to be coded and the predictive block isreferred to as a residual block or prediction errors. An inter-coded block is encodedaccording to a motion vector that points to a reference block in a reference frameforming the predictive block, and the residual block. The process of determiningthe motion vector is typically referred to as motion estimation. An intra coded blockis encoded according to an intra prediction mode and the residual block. For furthercompression, the residual block is transformed from the pixel domain to a transformdomain, e.g., frequency domain, resulting in residual transform coefficients, whichmay then be quantized. The quantized transform coefficients, initially arranged ina two-dimensional array, may be scanned to produce a one-dimensional vector oftransform coefficients, and then entropy encoded into a video bitstream to achieveeven more compression.
[0006] The encoded video bitstream is then saved in a computer-readable storagemedium (e.g., flash memory) to be accessed by another electronic device withdigital video capability or directly transmitted to the electronic device wired orwirelessly. The electronic device then performs video decompression (which is anopposite process to the video compression described above) by, e.g., parsing theencoded video bitstream to obtain syntax elements from the bitstream andreconstructing the digital video data to its original format from the encoded videobitstream based at least in part on the syntax elements obtained from the bitstream,and renders the reconstructed digital video data on a display of the electronicdevice.
[0007] With digital video quality going from high definition, to 4Kx2K or even8Kx4K, the amount of vide data to be encoded / decoded grows exponentially. It isa constant challenge in terms of how the video data may be encoded / decoded moreefficiently while maintaining the image quality of the decoded video data.
[0008] In a Joint Video Experts Team (JVET) meeting, JVET defined the firstdraft of Versatile Video Coding (VVC) and the VVC Test Model 1 (VTM1)encoding method. It was decided to include a quadtree with nested multi-type treeusing binary and ternary splits coding block structure as the initial new codingfeature of VVC. Since then, the reference software VTM to implement the encodingmethod and the draft VVC decoding process has been developed during the JVETmeetings.SUMMARY
[0009] In general, this disclosure describes examples of techniques relating tomotion compensated prediction using geometric shaped prediction unit in videocoding.
[0010] According to a first aspect of the present disclosure, there is provided amethod for video coding with geometric partition, including: partitioning videopictures into a plurality of coding units (CUs), at least one of which is furtherpartitioned into two prediction units (PUs), a first PU and a second PU, includingat least one geometric shaped PU; constructing a first merge list comprising aplurality of candidates, based on a merge list construction process for regular mergeprediction, wherein each one of the plurality of candidates is a motion vector (MV)comprising a List 0 MV, and / or a List 1 MV; locating a first candidate for the firstPU according to a first index; locating a second candidate for the second PUaccording to a second index; obtaining a first uni-prediction MV for the first PU byselecting a List X1 MV of the first candidate according to a first binary referencelist indication flag, wherein X1 takes a value of 0 or 1 and is indicated by the firstbinary reference list indication flag; and obtaining a second uni-prediction MV forthe second PU by selecting a List X2 MV of the second candidate according to asecond binary reference list indication flag, wherein X2 takes a value of 0 or 1 andis indicated by the second binary reference list indication flag.
[0011] According to a second aspect of the present disclosure, an apparatus forvideo coding with geometric partition is provided, including: one or moreprocessors; and a memory configured to store instructions executable by the one ormore processors; wherein the one or more processors, upon execution of theinstructions, are configured to: partition video pictures into a plurality of codingunits (CUs), at least one of which is further partitioned into two prediction units(PUs), a first PU and a second PU, including at least one geometric shaped PU;construct a first merge list comprising a plurality of candidates, based on a mergelist construction process for regular merge prediction wherein each one of theplurality of candidates is a motion vector (MV) comprising a List 0 MV, and / or aList 1 MV; locate a first candidate for the first PU according to a first index; locatea second candidate for the second PU according to a second index; obtain a firstuni-prediction MV for the first PU by selecting a List X1 MV of the first candidateaccording to a first binary reference list indication flag, wherein X1 takes a value of0 or 1 and is indicated by the first binary reference list indication flag; and obtain asecond uni-prediction MV for the second PU by selecting a List X2 MV of thesecond candidate according to a second binary reference list indication flag,wherein X2 takes a value of 0 or 1 and is indicated by the second binary referencelist indication flag.
[0012] According to a third aspect of the present disclosure, there is provided anon-transitory computer-readable storage medium for video coding with geometricpartition storing computer-executable instructions that, when executed by one ormore computer processors, cause the one or more computer processors to performacts, including: partitioning video pictures into a plurality of coding units (CUs), atleast one of which is further partitioned into two prediction units (PUs), a first PUand a second PU, including at least one geometric shaped PU; constructing a firstmerge list comprising a plurality of candidates, based on a merge list constructionprocess for regular merge prediction, wherein each one of the plurality of candidatesis a motion vector (MV) comprising a List 0 MV, and / or a List 1 MV; locating afirst candidate for the first PU according to a first index; locating a second candidatefor the second PU according to a second index; obtaining a first uni-prediction MVfor the first PU by selecting a List X1 MV of the first candidate according to a firstbinary reference list indication flag, wherein X1 takes a value of 0 or 1 and isindicated by the first binary reference list indication flag; and obtaining a seconduni-prediction MV for the second PU by selecting a List X2 MV of the secondcandidate according to a second binary reference list indication flag, wherein X2takes a value of 0 or 1 and is indicated by the second binary reference list indicationflag.BRIEF DESCRIPTION OF THE DRAWINGS
[0013] A more particular description of the examples of the present disclosurewill be rendered by reference to specific examples illustrated in the appendeddrawings. Given that these drawings depict only some examples and are nottherefore considered to be limiting in scope, the examples will be described andexplained with additional specificity and details through the use of theaccompanying drawings.
[0014] FIG. 1 is a block diagram illustrating an exemplary video encoder inaccordance with some implementations of the present disclosure.
[0015] FIG. 2 is a block diagram illustrating an exemplary video decoder inaccordance with some implementations of the present disclosure.
[0016] FIG. 3 is a schematic diagram illustrating a quadtree plus binary tree(QTBT) structure in accordance with some implementations of the presentdisclosure.
[0017] FIG. 4 is a schematic diagram illustrating an example of a picture dividedinto CTUs in accordance with some implementations of the present disclosure.
[0018] FIG. 5 is a schematic diagram illustrating multi-type tree splitting modesin accordance with some implementations of the present disclosure.
[0019] FIG. 6 is a schematic diagram illustrating splitting a CU into triangularprediction units in accordance with some implementations of the present disclosure.
[0020] FIG. 7 is a schematic diagram illustrating positions of neighboring blocksin accordance with some implementations of the present disclosure.
[0021] FIG. 8 is a schematic diagram illustrating examples of geometricpartitions in accordance with some implementations of the present disclosure.
[0022] FIG. 9 is a schematic diagram illustrating an example of geometricpartition representation in accordance with some implementations of the presentdisclosure.
[0023] FIG. 10 a schematic diagram illustrating positions of spatial mergecandidates in accordance with some implementations of the present disclosure.
[0024] FIG. 11 is a schematic diagram illustrating motion vector scaling for atemporal merge candidate in accordance with some implementations of the presentdisclosure.
[0025] FIG. 12 is a schematic diagram illustrating candidate positions for atemporal merge candidate in accordance with some implementations of the presentdisclosure.
[0026] FIG. 13A to FIG. 13B are schematic diagrams illustrating examples ofuni-prediction motion vector (MV) selection for triangle prediction mode inaccordance with some implementations of the present disclosure.
[0027] FIG. 14A to FIG. 14D are schematic diagrams illustrating examples ofuni-prediction MV selection for triangle prediction mode in accordance with someimplementations of the present disclosure.
[0028] FIG. 15A to FIG. 15B are schematic diagrams illustrating examples ofuni-prediction MV selection for triangle prediction mode in accordance with someimplementations of the present disclosure.
[0029] FIG. 16 is a schematic diagram illustrating an example of flexible uniprediction MV selection for triangle prediction mode in accordance with someimplementations of the present disclosure.
[0030] FIG. 17 is a block diagram illustrating an exemplary apparatus for videocoding in accordance with some implementations of the present disclosure.
[0031] FIG. 18 is a flowchart illustrating an exemplary process of video codingfor motion compensated prediction using geometric prediction unit in accordancewith some implementations of the present disclosure.DETAILED DESCRIPTION
[0032] Reference will now be made in detail to specific implementations,examples of which are illustrated in the accompanying drawings. In the followingdetailed description, numerous non-limiting specific details are set forth in order toassist in understanding the subject matter presented herein. But it will be apparentto one of ordinary skill in the art that various alternatives may be used. For example,it will be apparent to one of ordinary skill in the art that the subject matter presentedherein may be implemented on many types of electronic devices with digital videocapabilities.
[0033] Reference throughout this specification to "one embodiment," "anembodiment," "an example," "some embodiments," "some examples," or similarlanguage means that a particular feature, structure, or characteristic described isincluded in at least one embodiment or example. Features, structures, elements, orcharacteristics described in connection with one or some embodiments are alsoapplicable to other embodiments, unless expressly specified otherwise.
[0034] Throughout the disclosure, the terms "first," "second," "third," etc. are allused as nomenclature only for references to relevant elements, e.g. devices,components, compositions, steps, etc., without implying any spatial orchronological orders, unless expressly specified otherwise. For example, a "firstdevice" and a "second device" may refer to two separately formed devices, or twoparts, components or operational states of a same device, and may be namedarbitrarily.
[0035] As used herein, the term "if" or "when" may be understood to mean"upon" or "in response to" depending on the context. These terms, if appear in aclaim, may not indicate that the relevant limitations or features are conditional oroptional.
[0036] The terms "module," "sub-module," "circuit," "sub-circuit," "circuitry,""sub-circuitry," "unit," or "sub-unit" may include memory (shared, dedicated, orgroup) that stores code or instructions that may be executed by one or moreprocessors. A module may include one or more circuits with or without stored codeor instructions. The module or circuit may include one or more components that aredirectly or indirectly connected. These components may or may not be physicallyattached to, or located adjacent to, one another.
[0037] A unit or module may be implemented purely by software, purely byhardware, or by a combination of hardware and software. In a pure softwareimplementation, for example, the unit or module may include functionally relatedcode blocks or software components, that are directly or indirectly linked together,so as to perform a particular function.
[0038] FIG. 1 shows a block diagram illustrating an exemplary block-basedhybrid video encoder 100 which may be used in conjunction with many videocoding standards using block-based processing. In the encoder 100, a video frameis partitioned into a plurality of video blocks for processing. For each given videoblock, a prediction is formed based on either an inter prediction approach or an intraprediction approach. In inter prediction, one or more predictors are formed throughmotion estimation and motion compensation, based on pixels from previouslyreconstructed frames. In intra prediction, predictors are formed based onreconstructed pixels in a current frame. Through mode decision, a best predictormay be chosen to predict a current block.
[0039] A prediction residual, representing the difference between a current videoblock and its predictor, is sent to a Transform circuitry 102. Transform coefficientsare then sent from the Transform circuitry 102 to a Quantization circuitry 104 forentropy reduction. Quantized coefficients are then fed to an Entropy Codingcircuitry 106 to generate a compressed video bitstream. As shown in FIG. 1,prediction-related information 110 from an inter prediction circuitry and / or an IntraPrediction circuitry 112, such as video block partition info, motion vectors,reference picture index, and intra prediction mode, are also fed through the EntropyCoding circuitry 106 and saved into a compressed video bitstream 114.
[0040] In the encoder 100, decoder-related circuitries are also needed in order toreconstruct pixels for the purpose of prediction. First, a prediction residual isreconstructed through an Inverse Quantization 116 and an Inverse Transformcircuitry 118. This reconstructed prediction residual is combined with a BlockPredictor 120 to generate un-filtered reconstructed pixels for a current video block.
[0041] Spatial prediction (or "intra prediction") uses pixels from samples ofalready coded neighboring blocks (which are called reference samples) in the samevideo frame as the current video block to predict the current video block.
[0042] Temporal prediction (also referred to as "inter prediction") usesreconstructed pixels from already-coded video pictures to predict the current videoblock. Temporal prediction reduces temporal redundancy inherent in the videosignal. Temporal prediction signal for a given coding unit (CU) or coding block isusually signaled by one or more motion vectors (MVs) which indicate the amountand the direction of motion between the current CU and its temporal reference.Further, if multiple reference pictures are supported, one reference picture index isadditionally sent, which is used to identify from which reference picture in thereference picture store the temporal prediction signal comes.
[0043] After spatial and / or temporal prediction is performed, an intra / inter modedecision circuitry 121 in the encoder 100 chooses the best prediction mode, forexample based on the rate-distortion optimization method. The block predictor 120is then subtracted from the current video block; and the resulting prediction residualis de-correlated using the transform circuitry 102 and the quantization circuitry104.The resulting quantized residual coefficients are inverse quantized by the inversequantization circuitry 116 and inverse transformed by the inverse transformcircuitry 118 to form the reconstructed residual, which is then added back to theprediction block to form the reconstructed signal of the CU. Further in-loop filtering115, such as a deblocking filter, a sample adaptive offset (SAO), and / or an adaptivein-loop filter (ALF) may be applied on the reconstructed CU before it is put in thereference picture store of the picture buffer 117 and used to code future videoblocks. To form the output video bitstream 114, coding mode (inter or intra),prediction mode information, motion information, and quantized residualcoefficients are all sent to the entropy coding unit 106 to be further compressed andpacked to form the bit-stream.
[0044] For example, a deblocking filter is available in AVC, HEVC as well asthe now-current version of VVC. In HEVC, an additional in-loop filter called SAO(sample adaptive offset) is defined to further improve coding efficiency. In the nowcurrent version of the VVC standard, yet another in-loop filter called ALF (adaptiveloop filter) is being actively investigated, and it has a good chance of being includedin the final standard.
[0045] These in-loop filter operations are optional. Performing these operationshelps to improve coding efficiency and visual quality. They may also be turned offas a decision rendered by the encoder 100 to save computational complexity.
[0046] It should be noted that intra prediction is usually based on unfilteredreconstructed pixels, while inter prediction is based on filtered reconstructed pixelsif these filter options are turned on by the encoder 100.
[0047] FIG. 2 is a block diagram illustrating an exemplary block-based videodecoder 200 which may be used in conjunction with many video coding standards.This decoder 200 is similar to the reconstruction-related section residing in theencoder 100 of FIG. 1. In the decoder 200, an incoming video bitstream 201 is firstdecoded through an Entropy Decoding 202 to derive quantized coefficient levelsand prediction-related information. The quantized coefficient levels are thenprocessed through an Inverse Quantization 204 and an Inverse Transform 206 toobtain a reconstructed prediction residual. A block predictor mechanism,implemented in an Intra / inter Mode Selector 212, is configured to perform eitheran Intra Prediction 208, or a Motion Compensation 210, based on decodedprediction information. A set of unfiltered reconstructed pixels are obtained bysumming up the reconstructed prediction residual from the Inverse Transform 206and a predictive output generated by the block predictor mechanism, using asummer 214.
[0048] The reconstructed block may further go through an In-Loop Filter 209before it is stored in a Picture Buffer 213 which functions as a reference picturestore. The reconstructed video in the Picture Buffer 213 may be sent to drive adisplay device, as well as used to predict future video blocks. In situations wherethe In-Loop Filter 209 is turned on, a filtering operation is performed on thesereconstructed pixels to derive a final reconstructed Video Output 222.
[0049] Video coding / decoding standards mentioned above, such as VVC, JEM,HEVC, MPEG-4, Part 10, are conceptually similar. For example, they all use blockbased processing. Block partitioning schemes in some standards are elaboratedbelow.High Efficient Video Coding (HEVC)
[0050] HEVC is based on a hybrid block-based motion-compensated transformcoding architecture. The basic unit for compression is termed coding tree unit(CTU). The maximum CTU size is defined as up to 64 by 64 luma pixels, and twoblocks of 32 by 32 chroma pixels for 4:2:0 chroma format. Each CTU may containone coding unit (CU) or recursively split into four smaller CUs until the predefinedminimum CU size is reached. Each CU (also named leaf CU) contains one ormultiple prediction units (PUs) and a tree of transform units (TUs).
[0051] In general, except for monochrome content, a CTU may include one lumacoding tree block (CTB) and two corresponding chroma CTBs; a CU may includeone luma coding block (CB) and two corresponding chroma CBs; a PU may includeone luma prediction block (PB) and two corresponding chroma PBs; and a TU mayinclude one luma transform block (TB) and two corresponding chroma TBs.However, exceptions may occur because the minimum TB size is 4x4 for both lumaand chroma (i.e., no 2x2 chroma TB is supported for 4:2:0 color format) and eachintra chroma CB always has only one intra chroma PB regardless of the number ofintra luma PBs in the corresponding intra luma CB.
[0052] For an intra CU, the luma CB may be predicted by one or four luma PBs,and each of the two chroma CBs is always predicted by one chroma PB, where eachluma PB has one intra luma prediction mode and the two chroma PBs share oneintra chroma prediction mode. Moreover, for the intra CU, the TB size cannot belarger than the PB size. In each PB, the intra prediction is applied to predict samplesof each TB inside the PB from neighboring reconstructed samples of the TB. Foreach PB, in addition to 33 directional intra prediction modes, DC and planar modesare also supported to predict flat regions and gradually varying regions,respectively.
[0053] For each inter PU, one of three prediction modes including inter, skip, andmerge, may be selected. Generally speaking, a motion vector competition (MVC)scheme is introduced to select a motion candidate from a given candidate set thatincludes spatial and temporal motion candidates. Multiple references to the motionestimation allow finding the best reference in 2 possible reconstructed referencepicture lists (namely List 0 and List 1). For the inter mode (termed AMVP mode,where AMVP stands for advanced motion vector prediction), inter predictionindicators (List 0, List 1, or bi-directional prediction), reference indices, motioncandidate indices, motion vector differences (MVDs) and prediction residual aretransmitted. As for the skip mode and the merge mode, only merge indices aretransmitted, and the current PU inherits the inter prediction indicator, referenceindices, and motion vectors from a neighboring PU referred by the coded mergeindex. In the case of a skip coded CU, the residual signal is also omitted.Joint Exploration Test Model (JEM)
[0054] The Joint Exploration Test Model (JEM) is built up on top of the HEVCtest model. The basic encoding and decoding flowchart of HEVC is kept unchangedin the JEM; however, the design elements of most important modules, including themodules of block structure, intra and inter prediction, residue transform, loop filterand entropy coding, are somewhat modified and additional coding tools are added.The following new coding features are included in the JEM.
[0055] In HEVC, a CTU is split into CUs by using a quadtree structure denotedas coding tree to adapt to various local characteristics. The decision whether to codea picture area using inter-picture (temporal) or intra-picture (spatial) prediction ismade at the CU level. Each CU may be further split into one, two or four PUsaccording to the PU splitting type. Inside one PU, the same prediction process isapplied and the relevant information is transmitted to the decoder on a PU basis.After obtaining the residual block by applying the prediction process based on thePU splitting type, a CU may be partitioned into transform units (TUs) according toanother quadtree structure similar to the coding tree for the CU. One of key featuresof the HEVC structure is that it has the multiple partition conceptions includingCU, PU, and TU.
[0056] FIG. 3 is a schematic diagram illustrating a quadtree plus binary tree(QTBT) structure in accordance with some implementations of the presentdisclosure.
[0057] The QTBT structure removes the concepts of multiple partition types, i.e.,it removes the separation of the CU, PU and TU concepts, and supports moreflexibility for CU partition shapes. In the QTBT block structure, a CU may haveeither a square or rectangular shape. As shown in FIG. 3, a coding tree unit (CTU)is first partitioned by a quaternary tree (i.e., quadtree) structure. The quadtree leafnodes may be further partitioned by a binary tree structure. There are two splittingtypes in the binary tree splitting: symmetric horizontal splitting and symmetricvertical splitting. The binary tree leaf nodes are called coding units (CUs), and thatsegmentation is used for prediction and transform processing without any furtherpartitioning. This means that the CU, PU and TU have the same block size in theQTBT coding block structure. In the JEM, a CU sometimes consists of codingblocks (CBs) of different colour components, e.g., one CU contains one luma CBand two chroma CBs in the case of P and B slices of the 4:2:0 chroma format, andsometimes consists of a CB of a single component, e.g., one CU contains only oneluma CB or just two chroma CBs in the case of I slices.
[0058] The following parameters are defined for the QTBT partitioning scheme.- CTU size: the root node size of a quadtree, the same concept as in theHEVC;- MinQTSize: the minimum allowed quadtree leaf node size;- MaxBTSize: the maximum allowed binary tree root node size;- MaxBTDepth: the maximum allowed binary tree depth;- MinBTSize: the minimum allowed binary tree leaf node size.
[0059] In one example of the QTBT partitioning structure, the CTU size is set as128x128 luma samples with two corresponding 64x64 blocks of chroma samples(with a 4:2:0 chroma format), the MinQTSize is set as 16x16, the MaxBTSize isset as 64x64, the MinBTSize (for both width and height) is set as 4x4, and theMaxBTDepth is set as 4. The quadtree partitioning is applied to the CTU first togenerate quadtree leaf nodes. The quadtree leaf nodes may have a size from 16x16(i.e., the MinQTSize) to 128x128 (i.e., the CTU size). If the quadtree leaf node is128x128, it will not be further split by the binary tree since the size exceeds theMaxBTSize (i.e., 64x64). Otherwise, the quadtree leaf node could be furtherpartitioned by the binary tree. Therefore, the quadtree leaf node is also the root nodefor the binary tree and it has the binary tree depth as 0. When the binary tree depthreaches MaxBTDepth (i.e., 4), no further splitting is considered. When the binarytree node has a width equal to MinBTSize (i.e., 4), no further horizontal splitting isconsidered. Similarly, when the binary tree node has a height equal to MinBTSize,no further vertical splitting is considered. The leaf nodes of the binary tree arefurther processed by prediction and transform processing without any furtherpartitioning. In the JEM, the maximum CTU size is 256x256 luma samples.
[0060] An example of block partitioning by using the QTBT scheme, and thecorresponding tree representation are illustrated in FIG. 3. The solid lines indicatequadtree splitting and dotted lines indicate binary tree splitting. As shown in FIG.3, the coding tree unit (CTU) 300 is first partitioned by a quadtree structure, andthree of the four quadtree leaf nodes 302, 304, 306, 308 are further partitioned byeither a quadtree structure or a binary tree structure. For example, the quadtree leafnode 306 is further partitioned by quadtree splitting; the quadtree leaf node 304 isfurther partitioned into two leaf nodes 304a, 304b by binary tree splitting; and thequadtree leaf node 302 is also further partitioned by binary tree splitting. In eachsplitting (i.e., non-leaf) node of the binary tree, one flag is signaled to indicatewhich splitting type (i.e., horizontal or vertical) is used, where 0 indicates horizontalsplitting and 1 indicates vertical splitting. For example, for the quadtree leaf node304, 0 is signaled to indicate horizontal splitting, and for the quadtree leaf node 302,1 is signaled to indicate vertical splitting. For quadtree splitting, there is no need toindicate the splitting type since quadtree splitting always splits a block bothhorizontally and vertically to produce 4 sub-blocks with an equal size.
[0061] In addition, the QTBT scheme supports the ability for the luma andchroma to have a separate QTBT structure. Currently, for P and B slices, the lumaand chroma CTBs in one CTU share the same QTBT structure. However, for Islices, the luma CTB is partitioned into CUs by a QTBT structure, and the chromaCTBs are partitioned into chroma CUs by another QTBT structure. This means thata CU in an I slice consists of a coding block of the luma component or coding blocksof two chroma components, and a CU in a P or B slice consists of coding blocks ofall three colour components.Versatile Video Coding (VVC)
[0062] In a Joint Video Experts Team (JVET) meeting, the JVET defined the firstdraft of the Versatile Video Coding (VVC) and the VVC Test Model 1 (VTM1)encoding method. It was decided to include a quadtree with nested multi-type treeusing binary and ternary splits coding block structure as the initial new codingfeature of VVC.
[0063] In VVC, the picture partitioning structure divides the input video intoblocks called coding tree units (CTUs). A CTU is split using a quadtree with nestedmulti-type tree structure into coding units (CUs), with a leaf coding unit (CU)defining a region sharing the same prediction mode (e.g. intra or inter). Here, theterm "unit" defines a region of an image covering all components; the term "block"is used to define a region covering a particular component (e.g. luma), and maydiffer in spatial location when considering the chroma sampling format such as4:2:0.Partitioning of the picture into CTUs
[0064] FIG. 4 is a schematic diagram illustrating an example of a picture dividedinto CTUs in accordance with some implementations of the present disclosure.
[0065] In VVC, pictures are divided into a sequence of CTUs, and the CTUconcept is the same as that of the HEVC. For a picture that has three sample arrays,a CTU consists of an NxN block of luma samples together with two correspondingblocks of chroma samples. FIG. 4 shows the example of a picture 400 divided intoCTUs 402.
[0066] The maximum allowed size of the luma block in a CTU is specified to be128x128 (although the maximum size of the luma transform blocks is 64x64).Partitioning of the CTUs using a tree structure
[0067] FIG. 5 is a schematic diagram illustrating multi-type tree splitting modesin accordance with some implementations of the present disclosure.
[0068] In HEVC, a CTU is split into CUs by using a quaternary-tree structuredenoted as coding tree to adapt to various local characteristics. The decisionwhether to code a picture area using inter-picture (temporal) or intra-picture(spatial) prediction is made at the leaf CU level. Each leaf CU may be further splitinto one, two or four PUs according to the PU splitting type. Inside one PU, thesame prediction process is applied, and the relevant information is transmitted tothe decoder on a PU basis. After obtaining the residual block by applying theprediction process based on the PU splitting type, a leaf CU may be partitioned intotransform units (TUs) according to another quaternary-tree structure similar to thecoding tree for the CU. One of key feature of the HEVC structure is that it has themultiple partition conceptions including CU, PU, and TU.
[0069] In VVC, a quadtree with nested multi-type tree using binary and ternarysplits segmentation structure replaces the concepts of multiple partition unit types,i.e. it removes the separation of the CU, PU and TU concepts except as needed forCUs that have a size too large for the maximum transform length, and supportsmore flexibility for CU partition shapes. In the coding tree structure, a CU mayhave either a square or rectangular shape. A coding tree unit (CTU) is firstpartitioned by a quaternary tree (i.e., quadtree) structure. Then the quaternary treeleaf nodes may be further partitioned by a multi-type tree structure. As shown inFIG. 5, there are four splitting types in multi-type tree structure: vertical binarysplitting 502 (SPLIT_BT_VER), horizontal binary splitting 504(SPLIT_BT_HOR), vertical ternary splitting 506 (SPLIT_TT_VER), andhorizontal ternary splitting 508 (SPLIT_TT_HOR). The multi-type tree leaf nodesare called coding units (CUs), and unless the CU is too large for the maximumtransform length, this segmentation is used for prediction and transform processingwithout any further partitioning. This means that, in most cases, the CU, PU andTU have the same block size in the quadtree with nested multi-type tree codingblock structure. The exception occurs when the maximum supported transformlength is smaller than the width or height of the color component of the CU. InVTM1, a CU consists of coding blocks (CBs) of different color components, e.g.,one CU contains one luma CB and two chroma CBs (unless the video ismonochrome, i.e., having only one color component).Partitioning CUs into multiple prediction units
[0070] In VVC, for each CU partitioned based on the structure illustrated above,prediction of the block content may be performed either on the whole CU block orin a sub-block manner explained in the following paragraphs. The operation unit ofsuch prediction is called prediction unit (or PU).
[0071] In the case of intra prediction (or intra-frame prediction), usually the sizeof the PU is equal to the size of the CU. In other words, the prediction is performedon the whole CU block. For inter prediction (or inter-frame prediction), the size ofthe PU may be equal or less than the size of the CU. In other words, there are caseswhere a CU may be split into multiple PUs for prediction.
[0072] Some examples of having the PU size smaller than the CU size include anaffine prediction mode, an Advanced Temporal Level Motion Vector Prediction(ATMVP) mode, and a triangle prediction mode, etc.
[0073] Under the affine prediction mode, a CU may be split into multiple 4x4PUs for prediction. Motion vectors may be derived for each 4x4 PU and motioncompensation may be performed accordingly on the 4x4 PU. Under the ATMVPmode, a CU may be split into one or multiple 8x8 PUs for prediction. Motionvectors are derived for each 8x8 PU and motion compensation may be performedaccordingly on the 8x8 PU. Under the triangle prediction mode, a CU may be splitinto two triangular shape prediction units. Motion vectors are derived for each PUand motion compensation is performed accordingly. The triangle prediction modeis supported for inter prediction. More details of the triangle prediction mode areillustrated below.Triangle prediction mode
[0074] FIG. 6 is a schematic diagram illustrating splitting a CU into triangularprediction units in accordance with some implementations of the present disclosure.
[0075] The concept of the triangle prediction mode is to introduce triangularpartitions for motion compensated prediction. The triangle prediction mode mayalso be named the triangular prediction unit mode. As shown in FIG. 6, a CU 602or 604 is split into two triangular prediction units PU1 and PU2, in either thediagonal or the inverse diagonal direction (i.e., either splitting from top-left cornerto bottom-right corner as shown in CU 602, or splitting from top-right corner tobottom-left corner as shown in CU 604). Each triangular prediction unit in the CUis inter-predicted using its own uni-prediction motion vector and reference frameindex which are derived from a uni-prediction candidate list. An adaptive weightingprocess is performed to the diagonal edge after predicting the triangular predictionunits. Then, the transform and quantization process are applied to the whole CU. Itis noted that this mode is only applied to skip and merge modes in the current VVC.Although in FIG. 6, the CU is shown as a square block, the triangle prediction modemay be applied to non-square (i.e. rectangular) shape CUs as well.
[0076] The uni-prediction candidate list may comprise one or more candidates,and each candidate may be a motion vector. Thus, throughout this disclosure, theterms "uni-prediction candidate list," "uni-prediction motion vector candidate list,"and "uni-prediction merge list" may be used interchangeably; and the terms "uniprediction merge candidates" and "uni-prediction motion vectors" may also be usedinterchangeably.Uni-prediction motion vector candidate list
[0077] FIG. 7 is a schematic diagram illustrating positions of the neighboringblocks in accordance with some implementations of the present disclosure.
[0078] In some examples, the uni-prediction motion vector candidate list mayinclude two to five uni-prediction motion vector candidates. In some otherexamples, other number may also be possible. It is derived from neighboringblocks. The uni-prediction motion vector candidate list is derived from sevenneighboring blocks including five spatial neighboring blocks (1 to 5) and twotemporal co-located blocks (6 to 7), as shown in FIG. 7. The motion vectors of theseven neighboring blocks are collected into a first merge list. Then, a uni-predictioncandidate list is formed based on the first merge list motion vectors according to aspecific order. Based on the order, the uni-prediction motion vectors from the firstmerge list are put in the uni-prediction motion vector candidate list first, followedby reference picture List 0 or L0 motion vector of bi-prediction motion vectors, andthen reference picture List 1 or L1 motion vector of bi-prediction motion vectors,and then followed by the averaged motion vector of the L0 and L1 motion vectorsof bi-prediction motion vectors. At that point, if the number of candidates is stillless than a target number (which may be five), zero motion vectors are added to thelist to meet the target number.
[0079] A predictor is derived for each of the triangular PUs based on its motionvector. It is worth noting that the predictor derived covers a larger area than theactual triangular PU so that there is an overlapped area of the two predictors alongthe shared diagonal edge of the two triangular PUs. A weighting process is appliedto the diagonal edge area between the two predictors to derive a final prediction forthe CU. The weighting factors currently used for the luminance and thechrominance samples are {7 / 8, 6 / 8, 5 / 8, 4 / 8, 3 / 8, 2 / 8, 1 / 8} and {6 / 8, 4 / 8, 2 / 8},respectively.Triangle prediction mode syntax and signaling
[0080] Triangle prediction mode is signaled using a triangle prediction flag. Thetriangle prediction flag is signaled when a CU is coded in either skip mode or mergemode. For a given CU, if the triangle prediction flag has a value of 1, it means thatthe corresponding CU is coded using triangle prediction mode. Otherwise, the CUis coded using a prediction mode other than triangle prediction mode.
[0081] For example, the triangle prediction flag is conditionally signaled in eitherskip mode or merge mode. Firstly, a triangle prediction tool enable / disable flag issignaled in sequence parameter set (or SPS). Only if this triangle prediction toolenable / disable flag is true, the triangle prediction flag is signaled at CU level.Secondly, triangle prediction tool is only allowed in B-slice. So only in a B-slice,the triangle prediction flag is signaled at CU level. Thirdly, triangle prediction modeis signaled only for a CU with a size equal or larger than a certain threshold. If aCU has a size smaller than that threshold, triangle prediction flag is not signaled.Fourthly, triangle prediction mode may only be signaled for a CU only if that CUis not coded in sub-block merge mode which includes both affine mode andATMVP mode. In the four cases listed above, when the triangle prediction flag isnot signaled, it is inferred as 0 at the decoder side.
[0082] When triangle prediction flag is signaled, it is signaled using Contextadaptive binary arithmetic coding (CABAC) entropy coder with certain contexts.The contexts are formed based on the triangle prediction flag values of the top andthe left block to the current CU.
[0083] To code (i.e. either encoder or decode) a triangle prediction flag for acurrent block (or a current CU), triangle prediction flags from both the top and theleft block (or CU) are derived and their values are summed up. This results in threepossible contexts corresponding to the following cases:1) Both the left block and the top block have a triangle prediction flag of 0;2) Both the left block and the top block have a triangle prediction flag of 1;3) Otherwise.
[0084] Separate probabilities are maintained for each of the three contexts. Oncea context value is determined for a current block, the triangle prediction flag of thecurrent block is coded using the CABAC probability model corresponding to thatcontext value.
[0085] If the triangle prediction flag is true, a triangle partition orientation flag issignaled to indicate if the partition is orientated from the top-left corner to thebottom-right corner or from the top-right corner to the bottom-left corner.
[0086] In the case where triangle prediction mode is used for a CU, two indexvalues are signaled to indicate the index values of the first and the second uniprediction merge candidate respectively for triangle prediction. These two indexvalues are used to locate two merge candidates from the uni-prediction motionvector candidate list described above, for the first and second partition, respectively.For triangle prediction, the two index values are required to be different so that thetwo predictors of the two triangular partitions may be different from each other. Asa result, the first index value is signaled directly. To signal the second index value,if it is smaller than the first index value, its value is signaled directly. Otherwise, itsvalue is subtracted by 1 before being signaled to the decoder. At the decoder side,the first index is decoded and used directly. To decode the second index value, avalue denoted as "idx" is firstly decoded from CABAC engine. The second indexvalue would be equal to the value of idx if idx is smaller than the first index value.Otherwise, the second index value would be equal to (idx+1).Geometric merge mode
[0087] Under geometric merge mode (i.e. GEO), a CU may be partitioned intotwo prediction units (i.e., geometric PUs) according to geometric partition. FIG. 8illustrates some examples of geometric partition in accordance with someimplementations of the present disclosure. As shown in FIG. 8, a CU may bepartitioned into two PUs along a line not exactly diagonal, and the two PUs mayhave geometric shapes such as triangle, wedge, or trapezoid shapes. The geometricmerge mode may be considered as an extension of the triangle prediction mode, ora superset of the triangle prediction mode.
[0088] FIG. 9 is a schematic diagram illustrating an example of geometricpartition representation in accordance with some implementations of the presentdisclosure. As shown in FIG. 9, given a GEO partition, signaling of the partitionrequires signaling for the angle α, and a separation line displacement, denoted as ρ,relatively to the center of the block. Here, α represents a quantized angle which isbetween 0 and 360 degrees.
[0089] Similar to triangle prediction mode, under GEO partitioning only uni prediction is allowed for each of the two GEO partitions, so that the memorybandwidth requirement under this mode does not exceed that for bi-predictedblocks at the decoder side. Motion vector prediction for GEO partitioning is alignedwith the triangle prediction mode. Also similar to triangle prediction mode, ablending operation based on weighted averaging may be applied on the innerboundary between the two partitions in the geometric merge mode. Therefore,techniques relating to motion compensated prediction using triangular predictionunit may also be applied to motion compensated prediction using geometric shapedprediction unit.Regular merge mode motion vector candidate list
[0090] According to the current VVC, under the regular merge mode where awhole CU is predicted without splitting into more than one PU, the motion vectorcandidate list or the merge candidate list is constructed using a different procedurethan that for the triangle prediction mode.
[0091] Firstly, spatial motion vector candidates are selected based on motionvectors from neighboring blocks as indicated in FIG. 10, which is a schematicdiagram illustrating positions of spatial merge candidates in accordance with someimplementations of the present disclosure. In the derivation of spatial mergecandidates of a current block 1002, a maximum of four merge candidates areselected among candidates that are located in positions as depicted in FIG. 10.These candidates are selected according to certain order. One exemplar order ofderivation is A1 → B1 → B0 → A0 → (B2). The position B2 is considered only whenany PU of positions A1, B1, B0, A0 is not available or is intra coded. It should benoted that other different orders may also be used.
[0092] Next, a temporal merge candidate is derived. In the derivation of thetemporal merge candidate, a scaled motion vector is derived based on the co-locatedPU belonging to the picture which has the smallest Picture Order Count (POC)difference with the current picture within the given reference picture list. Thereference picture list to be used for derivation of the co-located PU is explicitlysignaled in the slice header. The scaled motion vector for the temporal mergecandidate is obtained as illustrated by the dotted line in FIG. 11 which illustratesmotion vector scaling for the temporal merge candidate in accordance with someimplementations of the present disclosure. The scaled motion vector for thetemporal merge candidate is scaled from the motion vector of the co-located PUcol_PU using the POC distances, tb and td, where tb is defined to be the POCdifference between the reference picture of the current picture curr_ref and thecurrent picture curr_pic and td is defined to be the POC difference between thereference picture of the co-located picture col_ref and the co-located picturecol_pic. The reference picture index of the temporal merge candidate is set equal tozero. A practical realization of the scaling process is described in the HEVC draftspecification. For a B-slice, two motion vectors, one for reference picture List 0 andthe other for reference picture List 1, are obtained and combined to make the bipredictive merge candidate.
[0093] FIG. 12 is a schematic diagram illustrating candidate positions for thetemporal merge candidate in accordance with some implementations of the presentdisclosure.
[0094] The position of co-located PU is selected between two candidatepositions, C3 and H, as depicted in FIG. 12. If the PU at position H is not available,or is intra coded, or is outside of the current CTU, position C3 is used for thederivation of the temporal merge candidate. Otherwise, position H is used for thederivation of the temporal merge candidate.
[0095] After inserting both spatial and temporal motion vectors into the mergecandidate list as described above, history-based merge candidates are added. Theso-called history-based merge candidates include those motion vectors frompreviously coded CUs, which are maintained in a separate motion vector list, andmanaged based on certain rules.
[0096] After inserting history-based candidates, if the merge candidate list is notfull, pairwise average motion vector candidates are further added into the list. Asits name indicates, this type of candidates is constructed by averaging candidatesalready in the current list. More specifically, based on a certain order or rule, twocandidates in the merge candidate list are taken each time and the average motionvector of the two candidates is appended to the current list.
[0097] After inserting pairwise average motion vectors, if the merge candidatelist is still not full, zero motion vectors will be added to make the list full.Using regular merge list construction process to construct a first merge list fortriangle prediction
[0098] The triangle prediction mode in the current VVC shares some similaritieswith the regular merge prediction mode, in its overall procedures in forming apredictor. For example, under both prediction modes, a merge list needs to beconstructed based on at least the current CU's neighboring spatial motion vectorsand the co-located motion vectors. At the same time, the triangle prediction modealso has some aspects that are different from the regular merge prediction mode.
[0099] For example, although a merge list needs to be constructed under both thetriangle prediction mode and the regular merge prediction mode, the detailedprocedures of constructing such a list are different.
[0100] These differences incur additional cost to codec implementation asadditional logics are needed. The procedures and logics of constructing a merge listmay be unified and shared between the triangle prediction mode and the regularmerge prediction mode.
[0101] In some examples, in forming the uni-directional prediction (also calleduni-prediction) merge list for the triangle prediction mode, before adding a newmotion vector into the merge list, the new motion vector is fully pruned againstthose motion vectors already in the list. In other words, the new motion vector iscompared with each motion vector that is already in the uni-prediction merge list,and is added into the list only when it is different from every motion vector in thatmerge list. Otherwise, the new motion vector is not added into the list.
[0102] According to some examples of the present disclosure, under the triangleprediction mode, the uni-directional prediction merge list is constructed from theregular merge mode motion vector candidate list, which may be referred to as aregular merge list.
[0103] More specifically, to construct a merge candidate list for the triangleprediction mode, a first merge list is firstly constructed based on the merge listconstruction process for the regular merge prediction. The first merge list includinga plurality of candidates, each being a motion vector. Then, the uni-directionalprediction merge list for the triangle prediction mode is further constructed usingthe motion vectors in the first merge list.
[0104] It should be noted that the first merge list constructed in this case maychoose a different list size than that for the general merge mode or regular mergemode. In one example of the present disclosure, the first merge list has the samesize as that for the general merge mode. In another example of the presentdisclosure, the first merge list constructed has a list size different from that for thegeneral merge mode.Constructing uni-directional prediction merge list from the first merge list
[0105] According to some examples of the present disclosure, the uni-directionalprediction merge list for triangle prediction mode may be constructed from the firstmerge list based on one of the following methods.
[0106] In an example of the present disclosure, to construct the uni-directionalprediction merge list, prediction List 0 motion vectors of the candidates in the firstmerge list are checked and selected into the uni-directional prediction merge listfirst. If the uni-directional prediction merge list is not full (e.g., the number ofcandidates in this list is still less than the target number) after this process,prediction List 1 motion vectors of the candidates in the first merge list are checkedand selected into the uni-directional prediction merge list. If the uni-directionalprediction merge list is still not full, prediction List 0 zero vectors are added intothe uni-directional prediction merge list. If the uni-directional prediction merge liststill not full, prediction List 1 zero vectors are added into the uni-directionalprediction merge list.
[0107] In another example of the present disclosure, for each candidate in the firstmerge list, its prediction List 0 motion vector and prediction List 1 motion vectorare added in an interleaving manner into the uni-directional prediction merge list.More specifically, for each candidate in the first merge list, if a candidate is a unidirectional prediction motion vector, it is added directly into the uni-directionalprediction merge list. Otherwise, if the candidate is a bi-directional predictionmotion vector in the first merge list, its prediction List 0 motion vector is first addedinto the uni-directional prediction merge list, followed by its prediction List 1motion vector. Once all motion vector candidates in the first merge list are checkedand added, but the uni-directional prediction merge list is not full yet, unidirectional prediction zero motion vectors may be added. For example, for eachreference frame index, a prediction List 0 zero motion vector and a prediction List1 zero motion vector may be separately added into the uni-directional predictionmerge list until the list is full.
[0108] In yet another example of the present disclosure, the uni-directionalprediction motion vectors from the first merge list are selected into the unidirectional prediction merge list first. If the uni-directional prediction merge list isnot full after this process, for each bi-directional prediction motion vectors in thefirst merge list, its prediction List 0 motion vector is first added into the unidirectional prediction merge list, followed by its prediction List 1 motion vector.After this process, if the uni-directional prediction merge list is not full yet, unidirectional prediction zero motion vectors may be added. For example, for eachreference frame index, a prediction List 0 zero motion vector and a prediction List1 zero motion vector may be separately added into the uni-directional predictionmerge list until the list is full.
[0109] In the descriptions above, when a uni-directional prediction motion vectoris added into the uni-directional prediction merge list, a motion vector pruningprocess may be performed to make sure that the new motion vector to be added isdifferent from those motion vectors already in the uni-directional prediction mergelist. Such motion vector pruning process may also be performed in a partial mannerfor lower complexity, e.g., checking the new motion vector to be added only againstsome but not all motion vectors already in the uni-directional prediction merge list.In an extreme case, no motion vector pruning (i.e., motion vector comparisonoperation) is performed in the process.Constructing uni-directional prediction merge list from the first merge list based onpicture prediction configuration
[0110] In some examples of the present disclosure, the uni-prediction merge listmay be constructed in an adaptive manner based on whether a current picture usesbackward prediction. For example, the uni-prediction merge list may be constructedusing different methods depending on whether a current picture uses backwardprediction. If the Picture Order Count (POC) values of all the reference pictures arenot greater than the current picture's POC value, it means that the current picturedoes not use backward prediction.
[0111] In an example of the present disclosure, when a current picture does notuse backward prediction, or upon determining that the current picture does not usebackward prediction, prediction List 0 motion vectors of the candidates in the firstmerge list are checked and selected into the uni-directional prediction merge listfirst, followed by prediction List 1 motion vectors of those candidates; and if theuni-directional prediction merge list is still not full, uni-prediction zero motionvectors may be added. Otherwise, if the current picture uses backward prediction,prediction List 0 and List 1 motion vectors of each candidate in the first merge listmay be checked and selected into the uni-directional prediction merge list in aninterleaving manner as described above, i.e., the prediction List 0 motion vector ofthe first candidate in the first merge list is added, followed by the prediction List 1motion vector of the first candidate, and then the prediction List 0 motion vector ofthe second candidate is added, followed by the prediction List 1 motion vector ofthe second candidate, and so on. At the end of the process, if the uni-directionalprediction merge list is still not full, uni-prediction zero vectors may be added.
[0112] In another example of the present disclosure, if a current picture does notuse backward prediction, prediction List 1 motion vectors of the candidates in thefirst merge list are checked and selected into the uni-directional prediction mergelist first, followed by prediction List 0 motion vectors of those candidates; and ifthe uni-directional prediction merge list is still not full, uni-prediction zero motionvectors may be added. Otherwise, if the current picture uses backward prediction,prediction List 0 and List 1 motion vectors of each candidate in the first merge listmay be checked and selected into the uni-directional prediction merge list in aninterleaving manner as described above, i.e., the prediction List 0 motion vector ofthe first candidate in the first merge list is added, followed by the prediction List 1motion vector of the first candidate, and then the prediction List 0 motion vector ofthe second candidate is added, followed by the prediction List 1 motion vector ofthe second candidate, and so on. At the end of the process, if the uni-directionalprediction merge list is still not full, uni-prediction zero vectors may be added.
[0113] In yet another example of the present disclosure, if a current picture doesnot use backward prediction, only prediction List 0 motion vectors of the candidatesin the first merge list are checked and selected into the uni-directional predictionmerge list first, and if the uni-directional prediction merge list is still not full, uniprediction zero motion vectors may be added. Otherwise, if the current picture usesbackward prediction, prediction List 0 and List 1 motion vectors of each candidatein the first merge list may be checked and selected into the uni-directionalprediction merge list in an interleaving manner as described above, i.e., theprediction List 0 motion vector of the first candidate in the first merge list is added,followed by the prediction List 1 motion vector of the first candidate, and then theprediction List 0 motion vector of the second candidate is added, followed by theprediction List 1 motion vector of the second candidate, and so on. At the end ofthe process, if the uni-directional prediction merge list is still not full, uni-predictionzero vectors may be added.
[0114] In still another example of the present disclosure, if a current picture doesnot use backward prediction, only prediction List 1 motion vectors of the candidatesin the first merge list are checked and selected into the uni-directional predictionmerge list first, and if the uni-directional prediction merge list is still not full, uniprediction zero motion vectors may be added. Otherwise, if the current picture usesbackward prediction, prediction List 0 and List 1 motion vectors of each candidatein the first merge list may be checked and selected into the uni-directionalprediction merge list in an interleaving manner as described above, i.e., theprediction List 0 motion vector of the first candidate in the first merge list is added,followed by the prediction List 1 motion vector of the first candidate, and then theprediction List 0 motion vector of the second candidate is added, followed by theprediction List 1 motion vector of the second candidate, and so on. At the end ofthe process, if the uni-directional prediction merge list is still not full, uni-predictionzero vectors may be added.
[0115] In the above examples, a uni-directional prediction merge list for triangleprediction is constructed by selecting motion vectors from the first merge list intothe uni-directional prediction merge list. However, in practice, the methods may beimplemented in different ways, with or without the uni-directional prediction (oruni-prediction) merge list being physically formed. In some examples, the firstmerge list may be used directly without physically creating a uni-directionalprediction merge list. For example, the List 0 and / or List 1 motion vectors of eachcandidate in the first merge list may be simply indexed based on a certain order andaccessed directly from the first merge list. That is, after constructing the first mergelist which includes a plurality of candidates, each being one or more motion vectors,based on a merge list construction process for regular merge prediction, a unidirectional prediction merge list is not constructed, but instead, an index listingincluding a plurality of reference indices, each reference index being a reference toa motion vector of a candidate in the first merge list, is constructed. It should benoted that an order of indexing may follow any of the selecting orders described inthe examples in which a uni-directional prediction merge list is constructed.
[0116] In one example of the present disclosure, when a current picture does notuse backward prediction, prediction List 0 motion vectors of the candidates in thefirst merge list are used as the uni-directional prediction merge candidates, indexedaccording to the same index order as they are in the first merge list. That is, upondetermining that the POC of the current picture is greater than each one of the POCsof the reference pictures, the reference indices are arranged according to a sameorder of List 0 motion vectors of the candidates in the first merge list. Otherwise, ifthe current picture uses backward prediction, List 0 and List 1 motion vectors ofeach candidate in the first merge list are used as the uni-directional prediction mergecandidates, indexed based on an interleaving manner, i.e., List 0 motion vector ofthe first candidate in the first merge list followed by List 1 motion vector of the firstcandidate, and then List 0 motion vector of the second candidate followed by List1 motion vector of the second candidate, and so on. That is, upon determining thatthe POC of the current picture is smaller than at least one of the POCs of thereference pictures, the reference indices are arranged according to an interleavingmanner of List 0 and List 1 motion vectors of each candidate in the first merge list,where the candidate is a bi-directional prediction motion vector. In the case wherea candidate in the first merge list is a uni-directional motion vector, a zero motionvector is indexed as the uni-directional prediction merge candidate following themotion vector of that candidate. This ensures that for the case where the currentpicture uses backward prediction, each candidate in the first merge list, regardlessit is a bi-directional or uni-directional prediction motion vector, provides two unidirectional motion vectors as the uni-directional prediction merge candidates.
[0117] In another example of the present disclosure, when a current picture doesnot use backward prediction, prediction List 0 motion vectors of the candidates inthe first merge list are used as the uni-directional prediction merge candidates,indexed according to the same index order as they are in the first merge list.Otherwise, if the current picture uses backward prediction, List 0 and List 1 motionvectors of each candidate in the first merge list are used as the uni-directionalprediction merge candidates, indexed based on an interleaving manner as describedabove, i.e., List 0 motion vector of the first candidate in the first merge list followedby List 1 motion vector of the first candidate, and then List 0 motion vector of thesecond candidate followed by list 1 motion vector of the second candidate, and soon. In the case where a candidate in the first merge list is a uni-directional motionvector, the motion vector plus certain motion offset is indexed as the uni-directionalprediction merge candidate following the motion vector of the candidate.
[0118] In the above processes, when checking a new motion vector to be addedinto the uni-directional prediction merge list, pruning may be performed fully, orpartially. When it is performed partially, it means that the new motion vector iscompared against some, but not all, of the motion vectors that are already in theuni-prediction merge list. In an extreme case, no motion vector pruning (i.e. motionvector comparison operation) is performed in the process.
[0119] The motion vector pruning may also be performed adaptively in formingthe uni-prediction merge list, based on whether the current picture uses backwardprediction or not. For example, for the examples of the present disclosure relatingto index listing construction based on picture prediction configuration, when thecurrent picture does not use backward prediction, motion vector pruning operationis performed, either fully or partially. When the current picture uses backwardprediction, motion vector pruning operation is not performed.Selecting uni-prediction merge candidates for triangle prediction mode
[0120] In addition to the abovementioned examples, other ways of uni-predictionmerge list construction or uni-prediction merge candidate selection are disclosed.
[0121] In one example of the present disclosure, once the first merge list for theregular merge mode is constructed, uni-prediction merge candidates may beselected for triangle prediction according to the following rules:
[0122] for a motion vector candidate in the first merge list, one and only one ofits List 0 or List 1 motion vector is used for triangle prediction;
[0123] for a given motion vector candidate in the first merge list, if its mergeindex value in the list is an even number, its List 0 motion vector is used for triangleprediction if it is available, and in the case that this motion vector candidate doesnot have a List 1 motion vector, its List 0 motion vector is used for triangleprediction; and
[0124] for a given motion vector candidate in the first merge list, if its mergeindex value in the list is an odd number, its List 1 motion vector is used for triangleprediction if it is available, and in the case that this motion vector candidate doesnot have a List 1 motion vector, its List 0 motion vector is used for triangleprediction.
[0125] FIG. 13A shows an example of uni-prediction motion vector (MV)selection (or uni-prediction merge candidate selection) for triangle prediction mode.In the example, the first N (e.g. N equals to 5) merge MV candidates derived in thefirst merge list are indexed from 0 to (N-1); and each row has two columns,representing the List 0 and the List 1 motion vector respectively for a candidate inthe first merge list. Each candidate in the list may be either uni-predicted or bipredicted. For a uni-predicted candidate, it has only a List 0 or a List 1 motionvector, but not both. For a bi-predicted candidate, it has both List 0 and List 1motion vectors. In FIG. 13A, for each merge index, the motion vectors marked with"x" are those motion vectors to be used first for triangle prediction if they areavailable. If a motion vector marked in "x" is not available, the unmarked motionvector corresponding to the same merge index is then to be used for triangleprediction.
[0126] The above concept may be extended to other examples. FIG. 13B showsanother example of uni-prediction motion vector (MV) selection for triangleprediction mode. According to FIG. 13B, the rules for selecting uni-predictionmerge candidates for triangle prediction are as follows:
[0127] for a motion vector candidate in the first merge list, one and only one ofits List 0 or List 1 motion vector is used for triangle prediction;
[0128] for a given motion vector candidate in the first merge list, if its mergeindex value in the list is an even number, its List 1 motion vector is used for triangleprediction if it is available, and in the case that this motion vector candidate doesnot have a List 1 motion vector, its List 0 motion vector is used for triangleprediction; and
[0129] for a given motion vector candidate in the first merge list, if its mergeindex value in the list is an odd number, its List 0 motion vector is used for triangleprediction if it is available, and in the case that this motion vector candidate doesnot have a List 0 motion vector, its List 1 motion vector is used for triangleprediction.
[0130] In some examples, other different orders may be defined and used forselecting uni-prediction merge candidates for triangle prediction from those motionvector candidates in the first merge list. More specifically, for a given motion vectorcandidate in the first merge list, the decision of whether its List 0 or List 1 motionvector is used first when available for triangle prediction does not have to bedependent on the parity of the candidate's index value in the first merge list asdescribed above. For examples, the following rules may also be used:
[0131] for a motion vector candidate in the first merge list, one and only one ofits List 0 or List 1 motion vector is used for triangle prediction;
[0132] based on a certain pre-defined pattern, for a number of motion vectorcandidates in the first merge list, their List 0 motion vector are used for triangleprediction if available, and in the case that a List 0 motion vector does not exist, thecorresponding List 1 motion vector is used for triangle prediction; and
[0133] based on the same pre-defined pattern, for the remaining motion vectorcandidates in the first merge list, their List 1 motion vector are used for triangleprediction if available, and in the case that a List 1 motion vector does not exist, thecorresponding List 0 motion vector is used for triangle prediction.
[0134] FIGS. 14A to 14D show some examples of the pre-defined patterns in uniprediction motion vector (MV) selection for triangle prediction mode. For eachmerge index, the motion vectors marked with "x" are those motion vectors usedfirst for triangle prediction if they are available. If a motion vector marked in "x" isnot available, the unmarked motion vector corresponding to the same merge indexis then used for triangle prediction.
[0135] In FIG. 14A, for the first three motion vector candidates in the first mergelist, their List 0 motion vectors are checked first. Only when a List 0 motion vectoris not available, the corresponding List 1 motion vector is used for triangleprediction. For the fourth and fifth motion vector candidates in the first merge list,their List 1 motion vectors are checked first. Only when a List 1 motion vector isnot available, the corresponding List 0 motion vector is used for triangle prediction.FIGS. 14B to 14D show three other patterns in selecting uni-prediction mergecandidates from the first merge list. The examples shown in the figures are notlimiting, and there exist further examples. For instance, the horizontally and / orvertically mirrored versions of those patterns shown in FIGS. 14A to 14D may alsobe used.
[0136] The concept illustrated by some of the examples may be used jointly withthat illustrated by some other examples in the disclosure. According to one exampleof the disclosure, for a given motion vector candidate in the first merge list, thedecision of whether its List 0 or List 1 motion vector is used first when availablefor triangle and / or geometric prediction may be additionally dependent on whetherthe current picture and / or slice uses backward prediction. FIG. 15A to FIG. 15B areschematic diagrams illustrating examples of uni-prediction MV selection fortriangle prediction mode in accordance with some implementations of the presentdisclosure. As shown in FIG. 15A, if a current picture and / or slice uses backwardprediction, the selection order between a candidate's List 0 and List 1 motion vectoris based on the parity of the candidate's index value in the first merge list. If thecurrent picture and / or slice does not use backward prediction, a candidate's List x(with x being 0 or 1) motion vector is preferred and selected first if it is available,and an example where x is equal to 0 is shown in FIG. 15B. Similar to otherexamples explained earlier, in the case where a uni-prediction motion vector is notavailable for a certain candidate in the first merge list, the corresponding uniprediction motion vector of the same candidate from the other reference list isselected. As a result, for each candidate in the first merge list, a uni-predictionmotion vector can be obtained and used for triangle and / or geometric predictionmode. The obtained uni-prediction motion vector shares the same index value asthe merge candidate in the first merge list.Flexible merge list construction and candidate index signaling for triangleprediction mode
[0137] In one example of the disclosure, once the first merge list for the regularmerge mode is constructed, a uni-prediction motion vector is selected for triangleprediction directly from that list. To indicate a certain List 0 or List 1 motion vectorthat is used for triangle prediction, first, an index value is signaled to indicate whichcandidate from the first merge list is chosen. Then, a binary reference list indicationflag (referred to as L0L1_flag) is signaled to indicate if the List 0 or the List 1motion vector of that chosen candidate from the first merge list is selected for thefirst partition of triangle prediction. The same signaling method is used to indicatea second List 0 or List 1 motion vector to be used for the second partition of triangleprediction. For example, the syntaxes signaled for a triangle-mode-coded CU mayinclude index1, L0L1_flag1, index2, L0L1_flag2. Here, index1 and index2 are themerge index values of the two candidates selected from the first merge list for thefirst and the second partitions, respectively. Index1 and index2 are used to locatethe candidates for the first partition and the second partition (i.e. a first candidateand a second candidate), respectively. L0L1_flag1 is the binary flag for the firstpartition to indicate if the List 0 or List 1 motion vector of the chosen candidatebased on index1 from the first merge list is selected. L0L1_flag2 is the binary flagfor the second partition to indicate if the List 0 or List 1 motion vector of the chosencandidate based on index2 from the first merge list is selected. The uni-predictionMV for the first partition is obtained by selecting a List X1 MV of the first candidateaccording to the first binary reference list indication flag L0L1_flag1, where X1takes a value of 0 or 1. The uni-prediction MV for the second partition is obtainedby selecting a List X2 MV of the second candidate according to the second binaryreference list indication flag L0L1_flag2, where X2 takes a value of 0 or 1.
[0138] FIG. 16 is a schematic diagram illustrating an example of flexible uniprediction MV selection for triangle prediction mode in accordance with someimplementations of the present disclosure. As shown in FIG. 16, every List 0 and / orList 1 motion vector indicated with a symbol "x" in a rectangular box may beindicated / signaled to the decoder for deriving prediction for the first partition, andevery List 0 and / or List 1 motion vector indicated with a symbol "x" in a rectangularbox may be indicated / signaled to the decoder for deriving prediction for the secondpartition under triangle prediction mode. As a result, the selection of uni-predictionmotion vector from the first merge list becomes very flexible. Given a first mergelist with a size of N candidates, up to 2N uni-prediction motion vectors may be usedfor each of the two triangular partitions. The two merge index values for the twopartitions under triangle prediction mode do not have to be different from eachother. In other words, they may take the same value. The index values are signaleddirectly without adjustment before signaling. More specifically, unlike what isdefined in the current VVC, the second index value is signaled to the decoderdirectly without performing any adjustment to the value prior to signaling.
[0139] In another example of the disclosure, when the two index values are thesame, the binary flag for the second partition, L0L1_flag2, does not have to besignaled. Instead, it is inferred as having the contrary value relative to the binaryflag for the first partition, L0L1_flag1. In other words, in this case, L0L1_flag2 maytake a value of (1-L0L1_flag1).
[0140] In yet another example of the disclosure, the binary reference listindication flags, i.e. L0L1_flag1 and L0L1_flag2, may be coded as CABAC contextbins. The context used for L0L1_flag1 may be separate from the context used forL0L1_flag2. The CABAC probability under each context may be initialized at thebeginning of a video sequence, and / or at the beginning of a picture, and / or at abeginning of a tile group. In this example, different context modeling methods maybe used for coding these two flags. In one method, the context model selection forthe second flag may be dependent on the value of the first flag. In other words, atleast two context models may be used for coding the second flag, with the contextmodel selected based on the value of the signaled first flag. In another method, theCABAC probability under each context model may be initialized differentlydepending on whether a current picture uses backward prediction or not. In the casewhere a picture uses backward prediction, the probability may be initialized higherfor the flag to indicate a certain list, e.g. List 0. In yet another method, the first flagmay be coded as a bypass bin, and only the second flag is coded as a context bin.The methods mentioned here may be used either separately or jointly.
[0141] In still another example of the disclosure, when a motion vector indicatedby a merge index value and the associated binary reference list indication flag, i.e.L0L1_flag, does not exist, a uni-prediction zero motion vector may be used instead.That is, a uni-prediction zero MV is selected as the uni-prediction MV for the firstpartition upon determining that the List X1 MV of the first candidate does not exist;and / or a uni-prediction zero MV is selected as the uni-prediction MV for the secondpartition upon determining that the List X2 MV of the second candidate does notexist.
[0142] In still another example of the disclosure, when a motion vector indicatedby a merge index value and the associated L0L1_flag does not exist, thecorresponding motion vector indicated by the same merge index value but from theother list, i.e. List (1-L0L1_flag), may be used instead. That is, the uni-predictionMV for the first partition is obtained by selecting a List (1-X1) MV of the firstcandidate upon determining that the List X1 MV of the first candidate does not exist;and / or the uni-prediction MV for the second partition is obtained by selecting a List(1-X2) MV of the second candidate upon determining that the List X2 MV of thesecond candidate does not exist.
[0143] In still another example of the disclosure, for a triangle-mode-coded CU,the second L0L1_flag (i.e. L0L1_flag2) associated with the second index (i.e.index2) is not signaled but always inferred. In this case, the index1, L0L1_flag1and index2 syntaxes still need to be signaled. In one method, L0L1_flag2 is inferredbased on the value of the L0L1_flag1 and whether a current picture uses backwardprediction. More specifically, for a triangle-mode-coded CU, if the current pictureuses backward prediction, the value of L0L1_flag2 is inferred to be the oppositebinary value of L0L1_flag1 (i.e. 1 - L0L1_flag1); if the current picture does not usebackward prediction, the value of L0L1_flag2 is inferred to be the same asL0L1_flag1. Additionally, if the current picture does not use backward prediction,it may be further enforced that the value of index2 is different from the value ofindex1 because both the two motion vectors (one for each triangle partition) arefrom the same prediction list. If index2 has a value equal to index1, it means thatthe same motion vector would be used for both triangle partitions, which is notuseful from coding efficiency perspective. In this case, in signaling the value ofindex2, a corresponding adjustment to the value of index2 prior to indexbinarization may be performed, which is the same as that in the current VVC designfor index2 signaling. For example, in the case where the actual value of index1 issmaller than that of index2, a CABAC binarization codeword corresponding to(index2 - 1) is used to signal the value of index2; otherwise, a CABAC binarizationcodeword corresponding to index2 is used to signal the value of index2. Based onthis example of the disclosure, alternatively, the enforcement of index2 having adifferent value from index1 may also be applied to the case when the current pictureuses backward prediction, together with the same index2 value adjustment forCABAC binarization.
[0144] In still another example of the disclosure, for a triangle-mode-coded CU,none of the L0L1_flags are signaled. Instead, they are both inferred. In this case,the index1 and index2 syntaxes still need to be signaled, representing the mergeindex values of the two candidates selected from the first merge list for the first andthe second partitions, respectively. Given a merge candidate index value, certainapproaches may be defined or used in determining if the List 0 motion vector or theList 1 motion vector of the corresponding merge candidate from the first list isselected for triangle mode prediction. In one approach, for index1, the patternshown in FIG. 13A is used in determining from which prediction list a motionvector of a merge candidate is selected for triangle mode prediction; and for index2,the pattern shown in FIG. 13B is used in determining from which prediction list amotion vector of a merge candidate is selected for triangle mode prediction. In otherwords, if index1 is an even value, List 0 motion vector of the candidate indicatedby index1 is selected, and if index1 is an odd value, List 1 motion vector of thecandidate indicated by index1 is selected. For index2, if it is an even value, List 1motion vector is selected, and if it is an odd value, List 0 motion vector is selected.In the case where a motion vector corresponding to certain prediction list does notexist, certain default motion vector may be used instead, e.g. a zero motion vector,or the corresponding motion vector from the other prediction list, etc. It is alsopossible that the pattern shown in FIG. 13B is used for index1 and that shown inFIG. 13A used for index2, in determining from which prediction list a motion vectorof a merge candidate is selected for triangle mode prediction. That is, L0L1_flagsmay be determined based on values of the index1 and index2.
[0145] Although the methods in this disclosure are illustrated using triangleprediction mode as an example, given that geometric merge mode can be consideredas an extension or superset of the triangle prediction mode, the methods illustratedin this disclosure are naturally applicable to geometric merge mode as well. All themerge list construction methods illustrated in this disclosure can be naturally usedfor geometric merge mode without any modification. Likewise, all the merge indexsignaling methods illustrated in this disclosure can be naturally used for geometricmerge mode without any modification. In one example, the same merge listconstruction process according to FIG. 13A is used for geometric merge mode. Inanother example, the same merge list construction process according to FIG. 13Bis used for geometric merge mode. Also, it is worth mentioning that a same mergelist construction process can be shared and used for both triangle prediction modeand geometric merge mode, if both modes exist in VVC.
[0146] It should be mentioned that although a first merge list containing 5 mergecandidates is used in some examples in this disclosure, in practice the size of thefirst merge list may be defined differently, e.g. 6 or 4, or some other values. Themethods illustrated in the examples are applicable to the cases when the first mergelist has a size other than 5.
[0147] In the above examples, motion vector pruning may be performed as well.Such pruning may be done fully, or partially. When it is performed partially, itmeans a new motion vector is compared against some, but not all, of the motionvectors that are already in the uni-prediction merge list. It may also mean that onlysome, but not all, new motion vectors need to be checked for pruning before usedas merge candidates for triangle prediction. One specific example is that only thesecond motion vector is checked against the first motion vector for pruning beforeit is used as a merge candidate for triangle prediction, while all other motion vectorsare not checked for pruning. In the extreme case, no motion vector pruning (i.e.motion vector comparison operation) is performed in the process.
[0148] FIG. 17 is a block diagram illustrating an apparatus for video coding inaccordance with some implementations of the present disclosure. The apparatus1700 may be a terminal, such as a mobile phone, a tablet computer, a digitalbroadcast terminal, a tablet device, or a personal digital assistant.
[0149] As shown in FIG. 17, the apparatus 1700 may include one or more of thefollowing components: a processing component 1702, a memory 1704, a powersupply component 1706, a multimedia component 1408, an audio component 1710,an input / output (I / O) interface 1712, a sensor component 1714, and acommunication component 1716.
[0150] The processing component 1702 usually controls overall operations of theapparatus 1700, such as operations relating to display, a telephone call, datacommunication, a camera operation and a recording operation. The processingcomponent 1702 may include one or more processors 1720 for executinginstructions to complete all or a part of steps of the above method. Further, theprocessing component 1702 may include one or more modules to facilitateinteraction between the processing component 1702 and other components. Forexample, the processing component 1702 may include a multimedia module tofacilitate the interaction between the multimedia component 1708 and theprocessing component 1702.
[0151] The memory 1704 is configured to store different types of data to supportoperations of the apparatus 1700. Examples of such data include instructions,contact data, phonebook data, messages, pictures, videos, and so on for anyapplication or method that operates on the apparatus 1700. The memory 1704 maybe implemented by any type of volatile or non-volatile storage devices or acombination thereof, and the memory 1704 may be a Static Random AccessMemory (SRAM), an Electrically Erasable Programmable Read-Only Memory(EEPROM), an Erasable Programmable Read-Only Memory (EPROM), aProgrammable Read-Only Memory (PROM), a Read-Only Memory (ROM), amagnetic memory, a flash memory, a magnetic disk or a compact disk.
[0152] The power supply component 1706 supplies power for differentcomponents of the apparatus 1700. The power supply component 1706 may includea power supply management system, one or more power supplies, and othercomponents associated with generating, managing and distributing power for theapparatus 1700.
[0153] The multimedia component 1708 includes a screen providing an outputinterface between the apparatus 1700 and a user. In some examples, the screen mayinclude a Liquid Crystal Display (LCD) and a Touch Panel (TP). If the screenincludes a touch panel, the screen may be implemented as a touch screen receivingan input signal from a user. The touch panel may include one or more touch sensorsfor sensing a touch, a slide and a gesture on the touch panel. The touch sensor maynot only sense a boundary of a touching or sliding actions, but also detect durationand pressure related to the touching or sliding operation. In some examples, themultimedia component 1708 may include a front camera and / or a rear camera.When the apparatus 1700 is in an operation mode, such as a shooting mode or avideo mode, the front camera and / or the rear camera may receive externalmultimedia data.
[0154] The audio component 1710 is configured to output and / or input an audiosignal. For example, the audio component 1710 includes a microphone (MIC).When the apparatus 1700 is in an operating mode, such as a call mode, a recordingmode and a voice recognition mode, the microphone is configured to receive anexternal audio signal. The received audio signal may be further stored in thememory 1704 or sent via the communication component 1716. In some examples,the audio component 1710 further includes a speaker for outputting an audio signal.
[0155] The I / O interface 1712 provides an interface between the processingcomponent 1702 and a peripheral interface module. The above peripheral interfacemodule may be a keyboard, a click wheel, a button, or the like. These buttons mayinclude but not limited to, a home button, a volume button, a start button and a lockbutton.
[0156] The sensor component 1714 includes one or more sensors for providing astate assessment in different aspects for the apparatus 1700. For example, the sensorcomponent 1714 may detect an on / off state of the apparatus 1700 and relativelocations of components. For example, the components are a display and a keypadof the apparatus 1700. The sensor component 1714 may also detect a positionchange of the apparatus 1700 or a component of the apparatus 1700, presence orabsence of a contact of a user on the apparatus 1700, an orientation oracceleration / deceleration of the apparatus 1700, and a temperature change ofapparatus 1700. The sensor component 1714 may include a proximity sensorconfigured to detect presence of a nearby object without any physical touch. Thesensor component 1714 may further include an optical sensor, such as a CMOS orCCD image sensor used in an imaging application. In some examples, the sensorcomponent 1714 may further include an acceleration sensor, a gyroscope sensor, amagnetic sensor, a pressure sensor, or a temperature sensor.
[0157] The communication component 1716 is configured to facilitate wired orwireless communication between the apparatus 1700 and other devices. Theapparatus 1700 may access a wireless network based on a communication standard,such as WiFi, 4G, or a combination thereof. In an example, the communicationcomponent 1716 receives a broadcast signal or broadcast related information froman external broadcast management system via a broadcast channel. In an example,the communication component 1716 may further include a Near FieldCommunication (NFC) module for promoting short-range communication. Forexample, the NFC module may be implemented based on Radio FrequencyIdentification (RFID) technology, infrared data association (IrDA) technology,Ultra-Wide Band (UWB) technology, Bluetooth (BT) technology and othertechnology.
[0158] In an example, the apparatus 1700 may be implemented by one or moreof Application Specific Integrated Circuits (ASIC), Digital Signal Processors(DSP), Digital Signal Processing Devices (DSPD), Programmable Logic Devices(PLD), Field Programmable Gate Arrays (FPGA), controllers, microcontrollers,microprocessors or other electronic elements to perform the above method.
[0159] A non-transitory computer readable storage medium may be, for example,a Hard Disk Drive (HDD), a Solid-State Drive (SSD), Flash memory, a HybridDrive or Solid-State Hybrid Drive (SSHD), a Read-Only Memory (ROM), aCompact Disc Read-Only Memory (CD-ROM), a magnetic tape, a floppy disk andetc.
[0160] FIG. 18 is a flowchart illustrating an exemplary process of video codingfor motion compensated prediction using geometric partition in accordance withsome implementations of the present disclosure.
[0161] In step 1802, the processor 1720 partitions video pictures into a pluralityof coding units (CUs), at least one of which is further partitioned into two predictionunits (PUs). The two PUs may include at least one geometric shaped PU. Forexample, the geometric shaped PU may include a pair of triangular shaped PUs, apair of wedge-shaped PUs, or other geometric shaped PUs.
[0162] In step 1804, the processor 1720 constructs a first merge list including aplurality of candidates, each candidate being a motion vector comprising a List 0motion vector and / or a List 1 motion vector. For example, the processor 1720 mayconstruct the first merge list based on a merge list construction process for regularmerge prediction. The processor 1720 may obtain the first merge list from otherelectronic devices or storage as well.
[0163] In step 1806, the processor 1720 locates a first candidate for the first PUaccording to a first index.
[0164] In step 1808, the processor 1720 locates a second candidate for the secondPU according to a second index.
[0165] In step 1810, the processor 1720 obtains a first uni-prediction MV for thefirst PU by selecting a List X1 MV of the first candidate according to a first binaryreference list indication flag, wherein X1 takes a value of 0 or 1 and is indicated bythe first binary reference list indication flag.
[0166] In step 1812, the processor 1720 obtains a second uni-prediction MV forthe second PU by selecting a List X2 MV of the second candidate according to asecond binary reference list indication flag, wherein X2 takes a value of 0 or 1 andis indicated by the second binary reference list indication flag.
[0167] In some examples, there is provided an apparatus for video coding. Theapparatus includes a processor 1720; and a memory 1704 configured to storeinstructions executable by the processor; where the processor, upon execution ofthe instructions, is configured to perform a method as illustrated in FIG. 18.
[0168] In some other examples, there is provided a non-transitory computerreadable storage medium 1704, having instructions stored therein. When theinstructions are executed by a processor 1720, the instructions cause the processorto perform a method as illustrated in FIG. 18.
[0169] The description of the present disclosure has been presented for purposesof illustration, and is not intended to be exhaustive or limited to the presentdisclosure. Many modifications, variations, and alternative implementations will beapparent to those of ordinary skill in the art having the benefit of the teachingspresented in the foregoing descriptions and the associated drawings.
[0170] The examples were chosen and described in order to explain the principlesof the disclosure, and to enable others skilled in the art to understand the disclosurefor various implementations and to best utilize the underlying principles andvarious implementations with various modifications as are suited to the particularuse contemplated. Therefore, it is to be understood that the scope of the disclosureis not to be limited to the specific examples of the implementations disclosed andthat modifications and other implementations are intended to be included within thescope of the present disclosure.
Claims
1. A method for video encoding with geometric partition, comprising: partitioning video pictures into a plurality of coding units (CUs), at least one of which is further partitioned into two prediction units (PUs), a first PU and a second PU, including at least one geometric shaped PU; constructing a first merge list comprising a plurality of candidates, based on a merge list construction process for regular merge prediction, wherein each one of the plurality of candidates is a motion vector (MV) comprises a List 0 MV, and / or a List 1 MV; locating a first candidate for the first PU according to a first index; locating a second candidate for the second PU according to a second index; obtaining a first uni-prediction MV for the first PU by determining a List X1 MV of the first candidate according to a first binary reference list indication flag, wherein X1 takes a value of 0 or 1 and is indicated by the first binary reference list indication flag; and obtaining a second uni-prediction MV for the second PU by determining a List X2 MV of the second candidate according to a second binary reference list indication flag, wherein X2 takes a value of 0 or 1 and is indicated by the second binary reference list indication flag.
2. The method for video encoding with geometric partition as claimed in claim 1, wherein the second binary reference list indication flag is determined to have a contrary value relative to the first binary reference list indication flag upon determining that the first and second indexes are the same.
3. The method for video encoding with geometric partition as claimed in claim 1, wherein a uni-prediction zero MV is selected as the first uni-prediction MV upon determining that the List X1 MV of the first candidate does not exist; or a uni-prediction zero MV is selected as the second uni-prediction MV upon determining that the List X2 MV of the second candidate does not exist.
4. The method for video encoding with geometric partition as claimed in claim 1, wherein the second binary reference list indication flag is determined to have a contrary value relative to the first binary reference list indication flag upon determining that backward prediction is used for a current picture; and the second binary reference list indication flag is determined to have a same value as the first binary reference list indication flag upon determining that backward prediction is not used for the current picture.
5. The method for video encoding with geometric partition as claimed in claim 1, wherein the first and second binary reference list indication flags are determined based on values of the first and second indexes, respectively.
6. The method for video encoding with geometric partition as claimed in claim 5, wherein the first and second binary reference list indication flags are coded as CABAC context bins.
7. The method for video encoding with geometric partition as claimed in claim 1, wherein the first and second binary reference list indication flags are determined based on parities of index values of the first and second candidates in the first merge list, respectively.
8. An apparatus for video encoding with geometric partition, comprising: one or more processors; and a memory coupled to the one or more processors; wherein the one or more processors are configured to: partition video pictures into a plurality of coding units (CUs), at least one of which is further partitioned into two prediction units (PUs), a first PU and a second PU, including at least one geometric shaped PU; construct a first merge list comprising a plurality of candidates, based on a merge list construction process for regular merge prediction, wherein each one of the plurality of candidates is a motion vector (MV) comprising a List 0 MV, and / or a List 1 MV; locate a first candidate for the first PU according to a first index; locate a second candidate for the second PU according to a second index; obtain a first uni-prediction MV for the first PU by determining a List X1 MV of the first candidate according to a first binary reference list indication flag, wherein X1 takes a value of 0 or 1 and is indicated by the first binary reference list indication flag; and obtain a second uni-prediction MV for the second PU by determining a List X2 MV of the second candidate according to a second binary reference list indication flag, wherein X2 takes a value of 0 or 1 and is indicated by the second binary reference list indication flag.
9. The apparatus for video encoding with geometric partition as claimed in claim 8, wherein the second binary reference list indication flag is determined to have a contrary value relative to the first binary reference list indication flag upon determining that the first and second indexes are the same.
10. The apparatus for video encoding with geometric partition as claimed in claim 8, wherein a uni-prediction zero MV is selected as the first uni-prediction MV upon determining that the List X1 MV of the first candidate does not exist; or a uni-prediction zero MV is selected as the second uni-prediction MV upon determining that the List X2 MV of the second candidate does not exist.
11. The apparatus for video encoding with geometric partition as claimed in claim 8, wherein the second binary reference list indication flag is determined to have a contrary value relative to the first binary reference list indication flag upon determining that backward prediction is used for a current picture; and the second binary reference list indication flag is determined to have a same value as the first binary reference list indication flag upon determining that backward prediction is not used for the current picture.
12. The apparatus for video encoding with geometric partition as claimed in claim 8, wherein the first and second binary reference list indication flags are determined based on values of the first and second indexes, respectively.
13. The apparatus for video encoding with geometric partition as claimed in claim 8, wherein the first and second binary reference list indication flags are coded as CABAC context bins.
14. The apparatus for video encoding with geometric partition as claimed in claim 8, wherein the first and second binary reference list indication flags are determined based on parities of index values of the first and second candidates in the first merge list, respectively.
15. A method for transmitting a bitstream generated by an image encoding device, wherein the image encoding device comprises: an apparatus for video encoding with geometric partition as claimed in any of claims 8-14.
16. A method of storing a bitstream, comprising: storing the bitstream on a digital storage medium, wherein the bitstream comprises data generated by the method according to any of claims 1-7.
