WO2019188942A1

WO2019188942A1 - Systems and methods for performing motion compensated prediction for video coding

Info

Publication number: WO2019188942A1
Application number: PCT/JP2019/012420
Authority: WO
Inventors: Byeongdoo CHOI; Kiran Mukesh MISRA; Frank Bossen; Christopher Andrew Segall; Jie Zhao
Original assignee: Sharp Kabushiki Kaisha
Priority date: 2018-03-30
Filing date: 2019-03-25
Publication date: 2019-10-03

Abstract

This disclosure relates to video coding and more particularly to techniques for determining a motion vector predictor (MVP) for coding video data. According to an aspect of an invention, the MVP is selected as the motion vector associated with the lowest cost for each location in a search pattern about one or more origin points. The cost is determined for a pair of corresponding blocks based on the absolute sum of differences of the corresponding blocks and a side distortion cost.

Description

SYSTEMS AND METHODS FOR PERFORMING MOTION COMPENSATED PREDICTION FOR VIDEO CODING

This disclosure relates to video coding and more particularly to techniques for performing motion compensated prediction.

Digital video capabilities can be incorporated into a wide range of devices, including digital televisions, laptop or desktop computers, tablet computers, digital recording devices, digital media players, video gaming devices, cellular telephones, including so-called smartphones, medical imaging devices, and the like. Digital video may be coded according to a video coding standard. Video coding standards may incorporate video compression techniques. Examples of video coding standards include ISO/IEC MPEG-4 Visual and ITU-T H.264 (also known as ISO/IEC MPEG-4 AVC) and High-Efficiency Video Coding (HEVC). HEVC is described in High Efficiency Video Coding (HEVC), Rec. ITU-T H.265, December 2016, which is incorporated by reference, and referred to herein as ITU-T H.265. Extensions and improvements for ITU-T H.265 are currently being considered for the development of next generation video coding standards. For example, the ITU-T Video Coding Experts Group (VCEG) and ISO/IEC (Moving Picture Experts Group (MPEG) (collectively referred to as the Joint Video Exploration Team (JVET)) are studying the potential need for standardization of future video coding technology with a compression capability that significantly exceeds that of the current HEVC standard. The Joint Exploration Model 7 (JEM 7), Algorithm Description of Joint Exploration Test Model 7 (JEM 7), ISO/IEC JTC1/SC29/WG11 Document: JVET-G1001, July 2017, Torino, IT, which is incorporated by reference herein, describes the coding features that are under coordinated test model study by the JVET as potentially enhancing video coding technology beyond the capabilities of ITU-T H.265. It should be noted that the coding features of JEM 7 are implemented in JEM reference software. As used herein, the term JEM may collectively refer to algorithms included in JEM 7 and implementations of JEM reference software.

Video compression techniques enable data requirements for storing and transmitting video data to be reduced. Video compression techniques may reduce data requirements by exploiting the inherent redundancies in a video sequence. Video compression techniques may sub-divide a video sequence into successively smaller portions (i.e., groups of frames within a video sequence, a frame within a group of frames, slices within a frame, coding tree units (e.g., macroblocks) within a slice, coding blocks within a coding tree unit, etc.). Intra prediction coding techniques (e.g., intra-picture (spatial)) and inter prediction techniques (i.e., inter-picture (temporal)) may be used to generate difference values between a unit of video data to be coded and a reference unit of video data. The difference values may be referred to as residual data. Residual data may be coded as quantized transform coefficients. Syntax elements may relate residual data and a reference coding unit (e.g., intra-prediction mode indices, motion vectors, and block vectors). Residual data and syntax elements may be entropy coded. Entropy encoded residual data and syntax elements may be included in a compliant bitstream.

In one example, a method of determining a motion vector predictor for coding video data, the method comprises for each location in a search pattern about one or more origin points, determining a cost for a pair of corresponding blocks, wherein the cost is determined based on the absolute sum of differences of the corresponding blocks and a side distortion cost and selecting a motion vector predictor as the motion vector associated with the lowest cost.

FIG. 1 is a conceptual diagram illustrating an example of a group of pictures coded according to a quad tree binary tree partitioning in accordance with one or more techniques of this disclosure. FIG. 2 is a conceptual diagram illustrating an example of a video component sampling format in accordance with one or more techniques of this disclosure. FIG. 3 is a conceptual diagram illustrating possible coding structures for a block of video data in accordance with one or more techniques of this disclosure. FIG. 4A is a conceptual diagrams illustrating examples of coding a block of video data in accordance with one or more techniques of this disclosure. FIG. 4B is a conceptual diagrams illustrating examples of coding a block of video data in accordance with one or more techniques of this disclosure. FIG. 5 is a conceptual diagram illustrating the position of neighboring video blocks for inclusion in a set of candidate for motion vector predictors in accordance with one or more techniques of this disclosure. FIG. 6 is a conceptual diagram illustrating the position neighboring video blocks for inclusion in a set of candidate motion vector predictors in accordance with one or more techniques of this disclosure. FIG. 7A is a conceptual diagrams illustrating the use of template matching for deriving motion information for a current video block in accordance with one or more techniques of this disclosure. FIG. 7B is a conceptual diagrams illustrating the use of template matching for deriving motion information for a current video block in accordance with one or more techniques of this disclosure. FIG. 8A is a conceptual diagrams illustrating the use of bilateral matching for deriving motion information for a current video block in accordance with one or more techniques of this disclosure. FIG. 8B is a conceptual diagrams illustrating the use of bilateral matching for deriving motion information for a current video block in accordance with one or more techniques of this disclosure. FIG. 9 is a block diagram illustrating an example of a system that may be configured to encode and decode video data according to one or more techniques of this disclosure. FIG. 10 is a block diagram illustrating an example of a video encoder that may be configured to encode video data according to one or more techniques of this disclosure. FIG. 11 is a conceptual diagram illustrating the use of bilateral matching for deriving motion information for a current video block according to one or more techniques of this disclosure. FIG. 12 is a conceptual diagram illustrating the use of bilateral matching for deriving motion information for a current video block according to one or more techniques of this disclosure. FIG. 13 is a conceptual diagram illustrating a hierarchical search for use with template or bilateral matching according to one or more techniques of this disclosure. FIG. 14 is a block diagram illustrating an example of a video decoder that may be configured to decode video data according to one or more techniques of this disclosure. FIG. 15A is a conceptual diagram illustrating a search range according to one or more techniques of this disclosure. FIG. 15B is a conceptual diagram illustrating a search range according to one or more techniques of this disclosure.

In general, this disclosure describes various techniques for coding video data. In particular, this disclosure describes techniques for motion compensated prediction. It should be noted that although techniques of this disclosure are described with respect to ITU-T H.264, ITU-T H.265, and JEM, the techniques of this disclosure are generally applicable to video coding. For example, the coding techniques described herein may be incorporated into video coding systems, (including video coding systems based on future video coding standards) including block structures, intra prediction techniques, inter prediction techniques, transform techniques, filtering techniques, and/or entropy coding techniques other than those included in ITU-T H.265 and JEM. Thus, reference to ITU-T H.264, ITU-T H.265, and/or JEM is for descriptive purposes and should not be construed to limit the scope of the techniques described herein. Further, it should be noted that incorporation by reference of documents herein is for descriptive purposes and should not be construed to limit or create ambiguity with respect to terms used herein. For example, in the case where an incorporated reference provides a different definition of a term than another incorporated reference and/or as the term is used herein, the term should be interpreted in a manner that broadly includes each respective definition and/or in a manner that includes each of the particular definitions in the alternative.

In one example, a device for reconstructing video data comprises one or more processors configured to for each location in a search pattern about one or more origin points, determine a cost for a pair of corresponding blocks, wherein the cost is determined based on the absolute sum of differences of the corresponding blocks and a side distortion cost and select a motion vector predictor as the motion vector associated with the lowest cost.

In one example, a non-transitory computer-readable storage medium comprises instructions stored thereon that, when executed, cause one or more processors of a device to for each location in a search pattern about one or more origin points, determine a cost for a pair of corresponding blocks, wherein the cost is determined based on the absolute sum of differences of the corresponding blocks and a side distortion cost and select a motion vector predictor as the motion vector associated with the lowest cost.

In one example, an apparatus comprises means for determining a cost for a pair of corresponding blocks for each location in a search pattern about one or more origin points, wherein the cost is determined based on the absolute sum of differences of the corresponding blocks and a side distortion cost and means for selecting a motion vector predictor as the motion vector associated with the lowest cost.

The details of one or more examples are set forth in the accompanying drawings and the description below. Other features, objects, and advantages will be apparent from the description and drawings, and from the claims.

Video content typically includes video sequences comprised of a series of frames (or pictures). A series of frames may also be referred to as a group of pictures (GOP). Each video frame or picture may include a plurality of slices or tiles, where a slice or tile includes a plurality of video blocks. As used herein, the term video block may generally refer to an area of a picture or may more specifically refer to the largest array of sample values that may be predictively coded, sub-divisions thereof, and/or corresponding structures. Further, the term current video block may refer to an area of a picture being encoded or decoded. A video block may be defined as an array of sample values that may be predictively coded. It should be noted that in some cases pixels values may be described as including sample values for respective components of video data, which may also be referred to as color components, (e.g., luma (Y) and chroma (Cb and Cr) components or red, green, and blue components). It should be noted that in some cases, the terms pixel value and sample value are used interchangeably. Video blocks may be ordered within a picture according to a scan pattern (e.g., a raster scan). A video encoder may perform predictive encoding on video blocks and sub-divisions thereof. Video blocks and sub-divisions thereof may be referred to as nodes.

ITU-T H.264 specifies a macroblock including 16x16 luma samples. That is, in ITU-T H.264, a picture is segmented into macroblocks. ITU-T H.265 specifies an analogous Coding Tree Unit (CTU) structure (also referred to as an largest coding unit (LCU)). In ITU-T H.265, pictures are segmented into CTUs. In ITU-T H.265, for a picture, a CTU size may be set as including 16x16, 32x32, or 64x64 luma samples. In ITU-T H.265, a CTU is composed of respective Coding Tree Blocks (CTB) for each component of video data (e.g., luma (Y) and chroma (Cb and Cr). Further, in ITU-T H.265, a CTU may be partitioned according to a quadtree (QT) partitioning structure, which results in the CTBs of the CTU being partitioned into Coding Blocks (CB). That is, in ITU-T H.265, a CTU may be partitioned into quadtree leaf nodes. According to ITU-T H.265, one luma CB together with two corresponding chroma CBs and associated syntax elements are referred to as a coding unit (CU). In ITU-T H.265, a minimum allowed size of a CB may be signaled. In ITU-T H.265, the smallest minimum allowed size of a luma CB is 8x8 luma samples. In ITU-T H.265, the decision to code a picture area using intra prediction or inter prediction is made at the CU level.

In ITU-T H.265, a CU is associated with a prediction unit (PU) structure having its root at the CU. In ITU-T H.265, PU structures allow luma and chroma CBs to be split for purposes of generating corresponding reference samples. That is, in ITU-T H.265, luma and chroma CBs may be split into respect luma and chroma prediction blocks (PBs), where a PB includes a block of sample values for which the same prediction is applied. In ITU-T H.265, a CB may be partitioned into 1, 2, or 4 PBs. ITU-T H.265 supports PB sizes from 64x64 samples down to 4x4 samples. In ITU-T H.265, square PBs are supported for intra prediction, where a CB may form the PB or the CB may be split into four square PBs (i.e., intra prediction PB types include MxM or M/2xM/2, where M is the height and width of the square CB). In ITU-T H.265, in addition to the square PBs, rectangular PBs are supported for inter prediction, where a CB may by halved vertically or horizontally to form PBs (i.e., inter prediction PB types include MxM, M/2xM/2, M/2xM, or MxM/2). Further, it should be noted that in ITU-T H.265, for inter prediction, four asymmetric PB partitions are supported, where the CB is partitioned into two PBs at one quarter of the height (at the top or the bottom) or width (at the left or the right) of the CB (i.e., asymmetric partitions include M/4xM left, M/4xM right, MxM/4 top, and MxM/4 bottom). Intra prediction data (e.g., intra prediction mode syntax elements) or inter prediction data (e.g., motion data syntax elements) corresponding to a PB is used to produce reference and/or predicted sample values for the PB.

JEM specifies a CTU having a maximum size of 256x256 luma samples. JEM specifies a quadtree plus binary tree (QTBT) block structure. In JEM, the QTBT structure enables quadtree leaf nodes to be further partitioned by a binary tree (BT) structure. That is, in JEM, the binary tree structure enables quadtree leaf nodes to be recursively divided vertically or horizontally. FIG. 1 illustrates an example of a CTU (e.g., a CTU having a size of 256x256 luma samples) being partitioned into quadtree leaf nodes and quadtree leaf nodes being further partitioned according to a binary tree. That is, in FIG. 1 dashed lines indicate additional binary tree partitions in a quadtree. Thus, the binary tree structure in JEM enables square and rectangular leaf nodes, where each leaf node includes a CB. As illustrated in FIG. 1, a picture included in a GOP may include slices, where each slice includes a sequence of CTUs and each CTU may be partitioned according to a QTBT structure. FIG. 1 illustrates an example of QTBT partitioning for one CTU included in a slice.

In JEM, a QTBT is signaled by signaling QT split flag and BT split mode syntax elements. Further, in JEM, luma and chroma components may have separate QTBT partitions. That is, in JEM luma and chroma components may be partitioned independently by signaling respective QTBTs. Currently, in JEM independent QTBT structures are enabled for slices using intra prediction techniques. Additionally, it should be noted that JEM includes the following parameters for signaling of a QTBT tree:

It should be noted that in some examples, MinQTSize, MaxBTSize, MaxBTDepth, and/or MinBTSize may be different for the different components of video. In JEM, CBs are used for prediction without any further partitioning. That is, in JEM, a CB may be a block of sample values on which the same prediction is applied. Thus, a JEM QTBT leaf node may be analogous a PB in ITU-T H.265.

A video sampling format, which may also be referred to as a chroma format, may define the number of chroma samples included in a CU with respect to the number of luma samples included in a CU. For example, for the 4:2:0 sampling format, the sampling rate for the luma component is twice that of the chroma components for both the horizontal and vertical directions. As a result, for a CU formatted according to the 4:2:0 format, the width and height of an array of samples for the luma component are twice that of each array of samples for the chroma components. FIG. 2 is a conceptual diagram illustrating an example of a coding unit formatted according to a 4:2:0 sample format. FIG. 2 illustrates the relative position of chroma samples with respect to luma samples within a CU. As described above, a CU is typically defined according to the number of horizontal and vertical luma samples. Thus, as illustrated in FIG. 2, a 16x16 CU formatted according to the 4:2:0 sample format includes 16x16 samples of luma components and 8x8 samples for each chroma component. Further, in the example illustrated in FIG. 2, the relative position of chroma samples with respect to luma samples for video blocks neighboring the 16x16 CU are illustrated. For a CU formatted according to the 4:2:2 format, the width of an array of samples for the luma component is twice that of the width of an array of samples for each chroma component, but the height of the array of samples for the luma component is equal to the height of an array of samples for each chroma component. Further, for a CU formatted according to the 4:4:4 format, an array of samples for the luma component has the same width and height as an array of samples for each chroma component.

As described above, intra prediction data or inter prediction data is used to produce reference sample values for a block of sample values. The difference between sample values included in a current PB, or another type of picture area structure, and associated reference samples (e.g., those generated using a prediction) may be referred to as residual data. Residual data may include respective arrays of difference values corresponding to each component of video data. Residual data may be in the pixel domain. A transform, such as, a discrete cosine transform (DCT), a discrete sine transform (DST), an integer transform, a wavelet transform, or a conceptually similar transform, may be applied to an array of difference values to generate transform coefficients. It should be noted that in ITU-T H.265, a CU is associated with a transform unit (TU) structure having its root at the CU level. That is, in ITU-T H.265, an array of difference values may be sub-divided for purposes of generating transform coefficients (e.g., four 8x8 transforms may be applied to a 16x16 array of residual values). For each component of video data, such sub-divisions of difference values may be referred to as Transform Blocks (TBs). It should be noted that in ITU-T H.265, TBs are not necessarily aligned with PBs. FIG. 3 illustrates examples of alternative PB and TB combinations that may be used for coding a particular CB. Further, it should be noted that in ITU-T H.265, TBs may have the following sizes 4x4, 8x8, 16x16, and 32x32. In JEM, residual values corresponding to a CB are used to generate transform coefficients without further partitioning. That is, in JEM a QTBT leaf node may be analogous to both a PB and a TB in ITU-T H.265. It should be noted that in JEM, a core transform and a subsequent secondary transforms may be applied (in the video encoder) to generate transform coefficients. For a video decoder, the order of transforms is reversed. Further, in JEM, whether a secondary transform is applied to generate transform coefficients may be dependent on a prediction mode.

A quantization process may be performed on transform coefficients. Quantization essentially scales transform coefficients in order to vary the amount of data required to represent a group of transform coefficients. Quantization may include division of transform coefficients by a quantization scaling factor and any associated rounding functions (e.g., rounding to the nearest integer). Quantized transform coefficients may be referred to as coefficient level values. Inverse quantization (or “dequantization”) may include multiplication of coefficient level values by the quantization scaling factor. It should be noted that as used herein the term quantization process in some instances may refer to division by a scaling factor to generate level values and multiplication by a scaling factor to recover transform coefficients in some instances. That is, a quantization process may refer to quantization in some cases and inverse quantization in some cases.

FIGS. 4A-4B are conceptual diagrams illustrating examples of coding a block of video data. As illustrated in FIG. 4A, a current block of video data (e.g., a CB corresponding to a video component) is encoded by generating a residual by subtracting a set of prediction values from the current block of video data, performing a transformation on the residual, and quantizing the transform coefficients to generate level values. As illustrated in FIG. 4B, the current block of video data is decoded by performing inverse quantization on level values, performing an inverse transform, and adding a set of prediction values to the resulting residual. It should be noted that in the examples in FIGS. 4A-4B, the sample values of the reconstructed block differs from the sample values of the current video block that is encoded. In this manner, coding may said to be lossy. However, the difference in sample values may be considered acceptable or imperceptible to a viewer of the reconstructed video. Further, as illustrated in FIGS. 4A-4B, scaling is performed using an array of scaling factors.

As illustrated in FIG. 4A, quantized transform coefficients are coded into a bitstream. Quantized transform coefficients and syntax elements (e.g., syntax elements indicating a coding structure for a video block) may be entropy coded according to an entropy coding technique. Examples of entropy coding techniques include content adaptive variable length coding (CAVLC), context adaptive binary arithmetic coding (CABAC), probability interval partitioning entropy coding (PIPE), and the like. Entropy encoded quantized transform coefficients and corresponding entropy encoded syntax elements may form a compliant bitstream that can be used to reproduce video data at a video decoder. An entropy coding process may include performing a binarization on syntax elements. Binarization refers to the process of converting a value of a syntax value into a series of one or more bits. These bits may be referred to as “bins.” Binarization is a lossless process and may include one or a combination of the following coding techniques: fixed length coding, unary coding, truncated unary coding, truncated Rice coding, Golomb coding, k-th order exponential Golomb coding, and Golomb-Rice coding. For example, binarization may include representing the integer value of 5 for a syntax element as 00000101 using an 8-bit fixed length binarization technique or representing the integer value of 5 as 11110 using a unary coding binarization technique. As used herein each of the terms fixed length coding, unary coding, truncated unary coding, truncated Rice coding, Golomb coding, k-th order exponential Golomb coding, and Golomb-Rice coding may refer to general implementations of these techniques and/or more specific implementations of these coding techniques. For example, a Golomb-Rice coding implementation may be specifically defined according to a video coding standard, for example, ITU-T H.265. An entropy coding process further includes coding bin values using lossless data compression algorithms. In the example of a CABAC, for a particular bin, a context model may be selected from a set of available context models associated with the bin. In some examples, a context model may be selected based on a previous bin and/or values of previous syntax elements. A context model may identify the probability of a bin having a particular value. For instance, a context model may indicate a 0.7 probability of coding a 0-valued bin and a 0.3 probability of coding a 1-valued bin. It should be noted that in some cases the probability of coding a 0-valued bin and probability of coding a 1-valued bin may not sum to 1. After selecting an available context model, a CABAC entropy encoder may arithmetically code a bin based on the identified context model. The context model may be updated based on the value of a coded bin. The context model may be updated based on an associated variable stored with the context, e.g., adaptation window size, number of bins coded using the context. It should be noted, that according to ITU-T H.265, a CABAC entropy encoder may be implemented, such that some syntax elements may be entropy encoded using arithmetic encoding without the usage of an explicitly assigned context model, such coding may be referred to as bypass coding.

As described above, intra prediction data or inter prediction data may associate an area of a picture (e.g., a PB or a CB) with corresponding reference samples. For intra prediction coding, an intra prediction mode may specify the location of reference samples within a picture. In ITU-T H.265, defined possible intra prediction modes include a planar (i.e., surface fitting) prediction mode (predMode: 0), a DC (i.e., flat overall averaging) prediction mode (predMode: 1), and 33 angular prediction modes (predMode: 2-34). In JEM, defined possible intra-prediction modes include a planar prediction mode (predMode: 0), a DC prediction mode (predMode: 1), and 65 angular prediction modes (predMode: 2-66). It should be noted that planar and DC prediction modes may be referred to as non-directional prediction modes and that angular prediction modes may be referred to as directional prediction modes. It should be noted that the techniques described herein may be generally applicable regardless of the number of defined possible prediction modes.

For inter prediction coding, a motion vector (MV) identifies reference samples in a picture other than the picture of a video block to be coded and thereby exploits temporal redundancy in video. For example, a current video block may be predicted from reference block(s) located in previously coded frame(s) and a motion vector may be used to indicate the location of the reference block. A motion vector and associated information may describe, for example, a horizontal component of the motion vector, a vertical component of the motion vector, a resolution for the motion vector (e.g., one-quarter pixel precision, one-half pixel precision, one-pixel precision, two-pixel precision, four-pixel precision), a prediction direction and/or a reference picture index value. Further, a coding standard, such as, for example ITU-T H.265, may support motion vector prediction. Motion vector prediction enables the value of a motion vector to be derived based on another motion vector. Examples of motion vector prediction include advanced motion vector prediction (AMVP), temporal motion vector prediction (TMVP), so-called “merge” mode, and “skip” and “direct” motion inference. Further, JEM supports advanced temporal motion vector prediction (ATMVP) and Spatial-temporal motion vector prediction (STMVP).

ITU-T H.265 supports two modes for motion vector prediction: a merge mode and so-called Advanced Motion Vector Prediction (AMVP). In ITU-T H.265, for both the merge mode and the AMVP for a current PB, a set of candidate blocks is derived. Both a video encoder and video decoder perform the same process to derive a set of candidates. Thus, for a current video block, the same set of candidates is generated during encoding and decoding. A candidate block includes a video block having associated motion information from which motion information used to generate a prediction for a current video block can be derived. For the merge mode in ITU-T H.265, all motion information (i.e., motion vector displacement values, reference picture indices, and reference picture lists) associated with a selected candidate is inherited as the motion information for the current PB. That is, at a video encoder, a candidate block is selected from the derived set of candidates and an index value included in the bitstream indicates the selected candidate and thus, indicates the motion information for the current PB. For AMVP in ITU-T H.265, the motion vector information for the selected candidate is used as a motion vector predictor for the motion vector of the current PB. That is, at a video encoder, a candidate block is selected from the derived set of candidates and an index value indicating the selected candidate and a delta value indicating the difference between the motion vector predictor and the motion vector for the current PB are included in the bitstream.

In ITU-T H.265, a set of candidate blocks may be derived from spatial neighboring blocks, and temporal blocks. Further, generated (or default) motion information may be used for motion vector prediction. Whether motion information used for motion vector prediction of a current PB includes motion information associated with spatial neighboring blocks, motion information associated with temporal blocks, or generated motion information is dependent on the number of candidates to be included in a set, whether temporal motion vector prediction is enabled, the availability of blocks, and/or whether motion information associated with blocks is redundant.

For the merge mode in ITU-T H.265, a maximum number of candidates that may be included in a set of candidate blocks may be set and signaled by a video encoder and may be up to five. Further, a video encoder may disable the use of temporal motion vector candidates (e.g., in order to reduce the amount memory resources needed to store motion information at a video decoder) and signal whether the use of temporal motion vector candidates is enabled or disabled for a picture. FIG. 5 illustrates the position of spatial neighboring blocks and the temporal block that may be included in a set of candidate blocks for the merge mode in ITU-T H.265. The derivation of the set of candidates for merge mode in ITU-T H.265 includes determining the availability of A1, B1, B0, A0, and B2. It should be noted that a block is considered unavailable, if it is intra-predicted (i.e., does not have corresponding motion information) or is not included in the current slice (or tile). After determining the availability of A1, B1, B0, A0, and B2, a set of comparisons (illustrated as dashed arrows in FIG. 5) is performed to remove redundant entries from the set of candidates. For example, B2 is compared to B1 and if B1 has associated motion information that is equal to that of B2, it is removed from the set of candidates. The removal of entries from a set of candidates may be referred to as a pruning process. It should be noted that in FIG. 5, in order to reduce complexity, a complete comparison of candidates is not performed (e.g., A0 is not compared to B0) and as such, it is possible that redundant entries are included in the set of candidates.

Referring again to FIG. 5, the dashed block labeled Temp refers to the temporal candidate that may be included in the set of candidates. In ITU-T H.265 for merge mode, for the temporal candidate, a spatially collocated PU included in a reference picture is defined and the temporal candidate includes a block having a position just outside to the bottom right of the collocated PU, if available, or the block at the center position of the collocated PU. As described above, a maximum number of candidates that may be included in a set of candidate blocks is set. If the maximum number of candidates is set to N, N-1 spatial candidates and the temporal candidate are included in the set, in cases where the number of available spatial candidates (after pruning) and temporal candidate is greater than or equal to N. In cases where the number of available spatial candidates (after pruning) and temporal candidate is less than N, generated motion information is included in the set in order to fill the set.

For the AMVP in ITU-T H.265, referring to FIG. 6, the derivation of the set of candidates includes adding one of A0 or A1 (i.e., a left candidate) and one of B0, B1 or B2 (an above candidate) to the set based on their availability. That is, the first available left candidate and the first available above candidate are added to the set. When the left candidate and the above candidate have redundant motion vector components, one redundant candidate is removed from the set. If the number of candidates included in the set is less than two, and temporal motion vector prediction is enabled, the temporal candidate (Temp) is included in the set. In cases where the number of available spatial candidates (after pruning) and temporal candidate included in the set is less than two, a zero value motion vector is included in the set in order to fill the set.

JEM describes a special merge mode, a so-called pattern matched motion vector derivation (PMMVD) mode, based on Frame-Rate Up Conversion (FRUC) techniques. In the PMMVD mode, both a video encoder and video decoder perform the same process to derive motion information of a video block. In some cases, in JEM a QTBT leaf node may be referred to as a CU. In JEM, a CU-level merge flag indicates whether motion vector information for the CU is derived using a merge mode. Further, in order to indicate that the PMMVD mode is to be used to derive motion information, when the merge flag is TRUE for a CU, a so-called FRUC flag is additionally signaled for the CU. When the FRUC flag is FALSE, a merge index is signaled and the regular merge mode is used to indicate motion information for the CU. When the FRUC flag is TRUE, a FRUC mode flag is additionally signaled to indicate whether one of a defined a template matching technique or a defined bilateral matching technique is to be used to derive motion information for sub-blocks within the CU. In JEM, each of the defined template matching technique and the defined bilateral matching technique derive an initial CU-level motion vector. After the initial CU-level motion vector is determined, the current CU is divided into sub-blocks and a motion vector is derived for each sub-block using the initial CU-level motion vector.

FIGS. 7A-7B illustrate the use of template matching for deriving a CU-level motion vector and subsequently a motion vector for a sub-block. In template matching, a template is defined, where a template indicates the positions of samples to be included in a set of samples. For a current video block, according to the template, a set of sample values in the current picture corresponding to the current video block are determined. For example, for an NxM current video block, a template may specify that for each of the N columns, the X samples neighboring the top row of the current video block and for each of the M rows, the Y samples neighboring the left column of the current video block are to be included in the set of samples. That is, for example, referring to FIG. 7A, for the current CU, a template may be an inverse L-shape including neighboring above and left samples.

Template matching is used to derive motion information for a current video block by determining the best match between the sample values in the current picture corresponding to the template and the current video block and a set of sample values positioned within a reference picture corresponding to the template. In JEM, for template matching the absolute sum difference (SAD) of corresponding samples is used to determine the degree to which two sets match, where lower values of the SAD indicate a better match. To search for sets of sample values positioned within a reference picture that may provide a good match, origin points and search patterns are used. Template matching techniques may use one or more sets of motion vector candidates to determine origin points. For example, the each of the motion vectors in the set of motion vector candidates {A1, B1, B0, A0, and B2} described above with respect to merge mode in ITU-T H.265 may be used as an origin point in a reference picture. For example, referring to FIG. 7A, for the current CU, the candidate motion vector B2 (i.e., MV_B2) is used as an origin point in reference picture. A region corresponding to an origin point may to searched for the set of samples that provides the best match. That is, a search may be performed according to a search pattern initiated with respect to an origin point. For example, if an origin point is (x₀, y₀) a search pattern may include evaluating sets of samples having the shape of the template located at the following positions: (x₀-4, y₀); (x₀+4, y₀); (x₀, y₀-4) and (x₀, y₀+4). In the example illustrated in FIG. 7A, MV_B2 forms the center of a search region and the set of samples at a position within the search region having the best match is indicated. As further illustrated in FIG. 7A, a potential CU-level motion vector (i.e., MV_CU) is determined from the relative position of the set of samples having the best match and the position of the current CU. The process illustrated in FIG. 7A, may be performed for each motion vector in the input set of candidate motion vectors and the CU-level motion vector for the current CU is selected from the potential CU-level motion vectors according to the potential CU-level motion vector providing the best match.

As described above, in JEM, after the initial CU-level motion vector is determined, the current CU is divided into sub-blocks and a motion vector is derived for each sub-block by performing a sub-block level iteration of template matching. The sub-block iteration of template matching uses the initial CU-level motion vector as input and may be referred to as a motion vector refinement process. FIG. 7B illustrates an example of motion vector refinement for motion vector MV_CU in FIG. 7A. Referring to FIG. 7B, for each sub-block, MV_CU is used as an origin point in the reference picture and a corresponding search region (i.e., search pattern) is used to determine the motion vector for each sub-block (i.e., for sub-block SB₀, the motion vector MV_SB0 is determined based on the best match).

It should be noted that in JEM, in cases where a candidate neighboring video block is predicted using bi-prediction (a first motion vector referencing a first reference picture and a second motion vector referencing a second reference picture), template matching is applied independently to a first reference picture and a second reference picture, and thereafter it is determined whether one of a Uni-prediction (one motion vector referencing one of the first reference picture or the second reference picture) or Bi-prediction (one motion vector referencing the first reference picture and one motion vector referencing the second reference picture) will be used to generate a prediction.

FIGS. 8A-8B illustrate the use of template matching for deriving a CU-level motion vector and subsequently a motion vector for a sub-block. Bilateral matching is used to derive motion information for a current video block by finding the closest match between two blocks in two different reference pictures located along a motion trajectory. A motion trajectory may include a motion vector, MV0, pointing to a reference block in a reference picture at a temporal distance, TD0, from a current picture, and a motion vector, MV1, pointing to a reference block in a reference picture at a temporal distance, TD1 from a current picture. For example, the reference picture at TD0 may be the temporal preceding picture to the current picture (which may be referred to as the backward picture) and the reference picture TD1 may be the temporal subsequent picture to the current picture (which may be referred to as the forward picture). It should be noted that in JEM, under the assumption of continuous motion trajectory, the motion vectors MV0 and MV1 pointing to the two reference blocks shall be proportional to the temporal distances, i.e., TD0 and TD1, between the current picture and the two reference pictures.

Similar to template matching, bilateral matching may use one or more sets of motion vector candidates at inputs and each motion vector candidate may be used as an origin point in a reference frame. For example, referring to FIG. 8A, for a current CU, the candidate motion vector B2 (i.e., MV_B2) is used as an origin point in the backward reference picture. Similar to template matching, a search may be performed according to a search pattern initiated with respect to an origin point to identify blocks of sample values in a reference picture. Further, in a manner similar to determining a potential CU-level motion vector in template matching described above, a potential CU-level backward prediction motion vector (i.e., MV0_CU in FIG. 8A) may be determined from the relative position of a block of sample value (i.e., Block₀ in FIG. 8A) and the position of the current CU. For each potential CU-level backward motion vector, a corresponding forward motion vector in generated. Typically, the corresponding forward motion vector is generated as a motion vector having the same absolute value and reverse direction of the backward motion vector and is further scaled proportionally based on the temporal distances of TD0 and TD1. The corresponding forward motion vector indicates a corresponding block of sample values (i.e., Block₁ in FIG. 8A). In JEM, the SAD of corresponding samples in Block₀ and Block₁ is used to determine the degree to which the two blocks match for determining the CU-level motion vectors MV0 and MV1. In JEM, for each origin point, potential motion vectors MV0 and MV1 are determined based on the best match between a block in the TD0 reference picture (e.g., Block₀ in FIG. 8A) and a block in the TD1 reference picture (e.g., Block₁ in FIG. 8A) within a search region. The initial CU-level motion vectors MV0 and MV1 are selected from the potential motion vectors MV0 and MV1 according to the best match of corresponding blocks.

As described above, in JEM, after the initial CU-level motion vector is determined, the current CU is divided into sub-blocks and a motion vector is derived for each sub-block by performing a sub-block level iteration of bilateral matching. FIG. 8B illustrates an example of motion vector refinement for MV0_CU and MV1_CU in FIG. 8A. Referring to FIG. 8B, for each sub-block, MV0_CU is used as an origin point in the backward reference picture and a corresponding search region (i.e., search pattern) is used to determine the motion vectors MV0 and MV1 for each sub-block (i.e., for sub-block SB₃, the motion vectors MV0_SB3 and MV1_SB3 are determined based on the best match). It should be noted that in JEM, the matching cost C of bilateral matching at sub-CU level search is calculated as follows:

For example, referring to FIG. 8B, MV is equal to MV0_CU and MV^s is equal to MV0_SB3.

The defined template matching technique and the defined bilateral matching technique in JEM may be less than ideal. For example, the techniques the bilateral matching technique in JEM may not yield reliable motion trajectories when objects have rotational or zooming motions, as opposed to simple translation motions. Further, the techniques in JEM for searching for a best match in template matching and bilateral matching may be inefficient.

FIG. 9 is a block diagram illustrating an example of a system that may be configured to code (i.e., encode and/or decode) video data according to one or more techniques of this disclosure. System 100 represents an example of a system that may perform video coding using motion vector prediction techniques described according to one or more examples of this disclosure. As illustrated in FIG. 9, system 100 includes source device 102, communications medium 110, and destination device 120. In the example illustrated in FIG. 9, source device 102 may include any device configured to encode video data and transmit encoded video data to communications medium 110. Destination device 120 may include any device configured to receive encoded video data via communications medium 110 and to decode encoded video data. Source device 102 and/or destination device 120 may include computing devices equipped for wired and/or wireless communications and may include set top boxes, digital video recorders, televisions, desktop, laptop, or tablet computers, gaming consoles, mobile devices, including, for example, “smart” phones, cellular telephones, personal gaming devices, and medical imagining devices.

Communications medium 110 may include any combination of wireless and wired communication media, and/or storage devices. Communications medium 110 may include coaxial cables, fiber optic cables, twisted pair cables, wireless transmitters and receivers, routers, switches, repeaters, base stations, or any other equipment that may be useful to facilitate communications between various devices and sites. Communications medium 110 may include one or more networks. For example, communications medium 110 may include a network configured to enable access to the World Wide Web, for example, the Internet. A network may operate according to a combination of one or more telecommunication protocols. Telecommunications protocols may include proprietary aspects and/or may include standardized telecommunication protocols. Examples of standardized telecommunications protocols include Digital Video Broadcasting (DVB) standards, Advanced Television Systems Committee (ATSC) standards, Integrated Services Digital Broadcasting (ISDB) standards, Data Over Cable Service Interface Specification (DOCSIS) standards, Global System Mobile Communications (GSM) standards, code division multiple access (CDMA) standards, 3rd Generation Partnership Project (3GPP) standards, European Telecommunications Standards Institute (ETSI) standards, Internet Protocol (IP) standards, Wireless Application Protocol (WAP) standards, and Institute of Electrical and Electronics Engineers (IEEE) standards.

Storage devices may include any type of device or storage medium capable of storing data. A storage medium may include a tangible or non-transitory computer-readable media. A computer readable medium may include optical discs, flash memory, magnetic memory, or any other suitable digital storage media. In some examples, a memory device or portions thereof may be described as non-volatile memory and in other examples portions of memory devices may be described as volatile memory. Examples of volatile memories may include random access memories (RAM), dynamic random access memories (DRAM), and static random access memories (SRAM). Examples of non-volatile memories may include magnetic hard discs, optical discs, floppy discs, flash memories, or forms of electrically programmable memories (EPROM) or electrically erasable and programmable (EEPROM) memories. Storage device(s) may include memory cards (e.g., a Secure Digital (SD) memory card), internal/external hard disk drives, and/or internal/external solid state drives. Data may be stored on a storage device according to a defined file format.

Referring again to FIG. 9, source device 102 includes video source 104, video encoder 106, and interface 108. Video source 104 may include any device configured to capture and/or store video data. For example, video source 104 may include a video camera and a storage device operably coupled thereto. Video encoder 106 may include any device configured to receive video data and generate a compliant bitstream representing the video data. A compliant bitstream may refer to a bitstream that a video decoder can receive and reproduce video data therefrom. Aspects of a compliant bitstream may be defined according to a video coding standard. When generating a compliant bitstream video encoder 106 may compress video data. Compression may be lossy (discernible or indiscernible) or lossless. Interface 108 may include any device configured to receive a compliant video bitstream and transmit and/or store the compliant video bitstream to a communications medium. Interface 108 may include a network interface card, such as an Ethernet card, and may include an optical transceiver, a radio frequency transceiver, or any other type of device that can send and/or receive information. Further, interface 108 may include a computer system interface that may enable a compliant video bitstream to be stored on a storage device. For example, interface 108 may include a chipset supporting Peripheral Component Interconnect (PCI) and Peripheral Component Interconnect Express (PCIe) bus protocols, proprietary bus protocols, Universal Serial Bus (USB) protocols, I²C, or any other logical and physical structure that may be used to interconnect peer devices.

Referring again to FIG. 9, destination device 120 includes interface 122, video decoder 124, and display 126. Interface 122 may include any device configured to receive a compliant video bitstream from a communications medium. Interface 108 may include a network interface card, such as an Ethernet card, and may include an optical transceiver, a radio frequency transceiver, or any other type of device that can receive and/or send information. Further, interface 122 may include a computer system interface enabling a compliant video bitstream to be retrieved from a storage device. For example, interface 122 may include a chipset supporting PCI and PCIe bus protocols, proprietary bus protocols, USB protocols, I²C, or any other logical and physical structure that may be used to interconnect peer devices. Video decoder 124 may include any device configured to receive a compliant bitstream and/or acceptable variations thereof and reproduce video data therefrom. Display 126 may include any device configured to display video data. Display 126 may comprise one of a variety of display devices such as a liquid crystal display (LCD), a plasma display, an organic light emitting diode (OLED) display, or another type of display. Display 126 may include a High Definition display or an Ultra High Definition display. It should be noted that although in the example illustrated in FIG. 9, video decoder 124 is described as outputting data to display 126, video decoder 124 may be configured to output video data to various types of devices and/or sub-components thereof. For example, video decoder 124 may be configured to output video data to any communication medium, as described herein.

FIG. 10 is a block diagram illustrating an example of video encoder 200 that may implement the techniques for encoding video data described herein. It should be noted that although example video encoder 200 is illustrated as having distinct functional blocks, such an illustration is for descriptive purposes and does not limit video encoder 200 and/or sub-components thereof to a particular hardware or software architecture. Functions of video encoder 200 may be realized using any combination of hardware, firmware, and/or software implementations. In one example, video encoder 200 may be configured to encode video data according to the techniques described herein. Video encoder 200 may perform intra prediction coding and inter prediction coding of picture areas, and, as such, may be referred to as a hybrid video encoder. In the example illustrated in FIG. 10, video encoder 200 receives source video blocks. In some examples, source video blocks may include areas of picture that has been divided according to a coding structure. For example, source video data may include macroblocks, CTUs, CBs, sub-divisions thereof, and/or another equivalent coding unit. In some examples, video encoder 200 may be configured to perform additional sub-divisions of source video blocks. It should be noted that some techniques described herein may be generally applicable to video coding, regardless of how source video data is partitioned prior to and/or during encoding. In the example illustrated in FIG. 10, video encoder 200 includes summer 202, transform coefficient generator 204, coefficient quantization unit 206, inverse quantization/transform processing unit 208, summer 210, intra prediction processing unit 212, inter prediction processing unit 214, filter unit 216, and entropy encoding unit 218.

As illustrated in FIG. 10, video encoder 200 receives source video blocks and outputs a bitstream. Video encoder 200 may generate residual data by subtracting a predictive video block from a source video block. Summer 202 represents a component configured to perform this subtraction operation. In one example, the subtraction of video blocks occurs in the pixel domain. Transform coefficient generator 204 applies a transform, such as a discrete cosine transform (DCT), a discrete sine transform (DST), or a conceptually similar transform, to the residual block or sub-divisions thereof (e.g., four 8x8 transforms may be applied to a 16x16 array of residual values) to produce a set of residual transform coefficients. Transform coefficient generator 204 may be configured to perform any and all combinations of the transforms included in the family of discrete trigonometric transforms. Transform coefficient generator 204 may output transform coefficients to coefficient quantization unit 206. Coefficient quantization unit 206 may be configured to perform quantization of the transform coefficients. As described above, the degree of quantization may be modified by adjusting a quantization parameter. Coefficient quantization unit 206 may be further configured to determine quantization parameters (QP) and output QP data (e.g., data used to determine a quantization group size and/or delta QP values) that may be used by a video decoder to reconstruct a quantization parameter to perform inverse quantization during video decoding. It should be noted that in other examples, one or more additional or alternative parameters may be used to determine a level of quantization (e.g., scaling factors). The techniques described herein may be generally applicable to determining a level of quantization for transform coefficients corresponding to a component of video data based on a level of quantization for transform coefficients corresponding another component of video data.

As illustrated in FIG. 10, quantized transform coefficients are output to inverse quantization/transform processing unit 208. Inverse quantization/transform processing unit 208 may be configured to apply an inverse quantization and an inverse transformation to generate reconstructed residual data. As illustrated in FIG. 10, at summer 210, reconstructed residual data may be added to a predictive video block. In this manner, an encoded video block may be reconstructed and the resulting reconstructed video block may be used to evaluate the encoding quality for a given prediction, transformation, and/or quantization. Video encoder 200 may be configured to perform multiple coding passes (e.g., perform encoding while varying one or more of a prediction, transformation parameters, and quantization parameters). The rate-distortion of a bitstream or other system parameters may be optimized based on evaluation of reconstructed video blocks. Further, reconstructed video blocks may be stored and used as reference for predicting subsequent blocks.

As described above, a video block may be coded using an intra prediction mode. Intra prediction processing unit 212 may be configured to select an intra prediction mode for a current video block. Intra prediction processing unit 212 may be configured to evaluate a frame and/or an area thereof and determine an intra prediction mode to use to encode a current block. As illustrated in FIG. 8, intra prediction processing unit 212 outputs intra prediction data (e.g., syntax elements) to entropy encoding unit 218 and transform coefficient generator 204. As described above, possible intra prediction modes may include planar prediction modes, DC prediction modes, and angular prediction modes. Inter prediction processing unit 214 may be configured to perform inter prediction coding for a current video block. Inter prediction processing unit 214 may be configured to receive source video blocks and calculate motion information for PUs of a video block. A motion vector may indicate the displacement of a PU (or similar coding structure) of a video block within a current video frame relative to a predictive block within a reference frame. Inter prediction coding may use one or more reference pictures. For example, inter prediction processing unit 214 may locate a predictive video block within a frame buffer (not shown in FIG. 10). It should be noted that inter prediction processing unit 214 may further be configured to apply one or more interpolation filters to a reconstructed residual block to calculate sub-integer pixel values for use in motion estimation. Further, motion prediction may be uni-predictive (use one motion vector) or bi-predictive (use two motion vectors). Inter prediction processing unit 214 may be configured to select a predictive block by calculating a pixel difference determined by, for example, sum of absolute difference (SAD), sum of square difference (SSD), or other difference metrics. Inter prediction processing unit 214 may output motion prediction data for a calculated motion vector to entropy encoding unit 218.

As described above, motion information may be determined and specified according to motion vector prediction techniques. Inter prediction processing unit 214 may be configured to perform motion vector prediction techniques, including for example, those described above. As described above, the template matching techniques and the bilateral matching techniques utilized in the PMMVD mode in JEM may be less than ideal. Inter prediction processing unit 214 may be configured to perform template matching and the bilateral matching according to the techniques described herein. As described above, in JEM, in bilateral matching the degree to which two blocks match is determined by at the CU-level by SAD and at the sub-block level by the matching cost C described above. In order to improve the robustness of bilateral matching, inter prediction processing unit 214 may be configured to determine a matching cost according to the following equation:

It should be noted that CBL is a bilateral matching cost according JEM, when β is zero the bilateral cost is determined in the same manner as JEM. In one example, β may be in the range of 0 to 1. In some examples, β may be a predefined value. Further, in some example, video encoder 200 may be configured to signal the value of β, for example, in a parameter set. In one example, the side distortion cost may be determined based on a side template match distortion (STD) and in one example, the side distortion cost may be determined based on a side continuity distortion (SCD). In one example, video encoder 200 may be configured to signal the method used to determine the side distortion cost.

Further, it should be noted that in some cases, the matching function may be modified as follows:

In this manner, a bilateral matching cost (e.g., the SAD of Block0 and Block1) and a side distortion cost can be relatively weighted.

The STD essentially uses a template matching technique in combination with bilateral matching. That is, for current video block a template is defined. In one example, the shape and size of the template may be the same as the template used in template matching in JEM. In another example, the shape and or size of the template signaled (or inferred) as having a size and shape other than that of the template specified in JEM. For example, in one example, the template shape, size, and location with respect to a current video block may be signaled video encode 200 in a parameter set. Referring to FIG. 11, for a current CU, a template is defined, based on the template neighboring samples corresponding to blocks in the backward picture and the forward picture are identify and as such an overall match cost may be determined based on the degree to which Block0 and Block1 match (e.g., using SAD, as in JEM) and the degree to which the neighboring samples match (which is weighted according to β). In one example, the degree to which neighboring samples match may be determined based on the following equation:

It should be noted that in some examples, in the equation above, the difference of the squared values of b0k and b1k may be used. It should be noted that in the STD equation above the cost function value is normalized by dividing the error value by the number of pixels. Further, calculating STD may cause blockiness to be reduced.

SCD measures the spatial smoothness across block boundaries of a block in a reference picture by computing the average of the absolute sample value differences between the boundary sample values in the block in a reference picture and corresponding reconstructed sample values in neighboring blocks. For example, referring to FIG. 12, for Block0, a degree of relative smoothness across its block boundaries may be determined by evaluating the differences between the boundary sample values in Block0 and the indicated neighboring samples. In one example, the degree to which neighboring samples match may be determined based on the following equation:

According to the equations provided above for C_R, the overall cost for a particular motion vector in SCD may be expressed as follows:

As such, a motion vector providing the minimum value of CSCD may be selected as the motion vector providing the best match.

As described above, in JEM, after the initial CU-level motion vector is determined, the current CU is divided into sub-blocks and a motion vector is derived for each sub-block using the initial CU-level motion vector. In one example, according to the techniques herein, in order to increase the accuracy of MV refinement for sub-blocks, the previously predicted image in neighboring sub-block through template matching or side template matching in bilateral matching is compared with the currently predicted image in the template region. In one example, when the MV of the current sub-block is re-searched, the predicted image of the neighboring(top & left) sub-block is recursively used. Thus, the neighboring (top & left) image is reconstructed by using the motion compensated prediction with the estimated MVs through the sub-block refinement for the neighboring sub-blocks. The cost value is obtained by comparing the reconstructed image and the motion compensated image using the current MV candidate.

As described above, in the bilateral matching in JEM, and as illustrated in the example in FIG. 8A, the paired motion vectors MV0CU and MV1CU have the exactly reverse direction with the same absolute value (or in some cases, a value scaled proportional to a picture order count (POC) difference). So, in JEM the forward motion vector and the corresponding backward motion vector are always symmetric. That is, in JEM, if the a forward motion vector is equal to (Vx + offset_x, Vy + offset_y), where (Vx, Vy) indicates the origin point and offset_x and offset_y indicate a shift in position from the origin (e.g., based on a search pattern), its corresponding backward motion vector shall be equal to (-Vx - offset_x, -Vy - offset_y).

In some cases, allowing the paired motion vectors, MV0 and MV1 to be asymmetric may result in improvements in coding efficiency. In one example, the asymmetric motion vectors may be derived using motion information of neighboring (spatial/temporal) blocks (e.g., from motion vectors of candidates in merge list belonging to respective reference picture list). In one example, according to the techniques herein, inter prediction processing unit 214 may be configured to generate paired motion vectors that are asymmetric. In one example, according to the techniques herein, a backward motion vector may have the value (V0x - offset_x, V0y - offset_y), in the case where the forward motion vector is equal to (V1x + offset_x, V1y + offset_y), where (V0x, V0y) and (V1x, V1y) indicate respective origin points. That is, according to the techniques described herein, a forward motion vector corresponding to an origin point (e.g., MVB2) may be paired with a backward motion vector corresponding to a distinct origin point (e.g., MVA1). In one example, the offsets offset_x and offset_y may be scaled based on the temporal distances, between the current picture and the two reference picture.

It should be noted that in JEM, in cases where a candidate neighboring video block is predicted using bi-prediction (a first motion vector referencing a first reference picture and a second motion vector referencing a second reference picture), bilateral matching is performed by independently performing bilateral matching with each of the motion vector as an origin point and the results are compared and the best match is selected. In one example, according to the techniques herein, in the case where a candidate neighboring video block is predicted using bi-prediction (e.g., a first and a second motion vector are associated with a video block), bilateral matching may be performed by setting a forward motion vector origin point to the first motion vector and setting a backward motion vector origin point to the second motion vector (or vice versa). It should be noted that in some examples, a search pattern may be defined such that blocks are compared at symmetric offsets. For example, the motion vector pair may have the following values (V0x - offset, V0y - offset) and (V1x + offset, V1y + offset), where (V0x, V0y) and (V1x, V1y) are the origin points. In one example, the offset values can be scaled proportionally according to temporal distance and/or the direction of reference frames. Furthermore, the offset values may be added to each origin point or subtracted from each origin point, for example, when the two references frames are located in the same temporal direction. Further, it should be noted that in some cases, an origin point based on motion vectors and offsets may comprise a pair of scalar values, where a first value represents displacement in the horizontal direction and a second value represents displacement in the vertical direction. In one example, the sign of an offset for a direction may be based on signs used in MV0 and MV1. For example, if MV0 and MV1 have the same sign, for example along the horizontal direction, then the offset will have same sign in the horizontal direction. Furthermore, opposite sign values may be used for offsets, for example, when the corresponding signs of MV0 and MV1 do not match.

As described above, the offsets, offset_x and offset_y may be scaled based on the temporal distances. It should be noted that in ITU-T H.265 and JEM, each picture is assigned a picture order count (POC) value indicating a temporal position of the picture and temporal distances may be determined as the difference between POC values. In one example, an offset for MV1, Offset1, may be calculated by scaling an offset for MV0, Offset0. For example, Offset1, may be calculated according following formula:

In this case, the motion vector pair may have the following values (V0x - offset0, V0y - offset0) and (V1x + offset1, V1y + offset1), where (V0x, V0y) and (V1x, V1y) are the origin points.

It should be noted that according to the techniques herein, symmetric bilateral matching or asymmetric bilateral matching can be selectively applied. In one example, a flag (e.g., signaled in a parameter set, a slice header, at the CU-level, etc.) can indicate which one is used. In one example, both types of matching may be performed and after both matchings are performed, the best one providing a lower RD cost may be selected. In one example, depending on one or more of a temporal Id in hierarchical B-structure, CU size, a partitioning depth, QP, variance of neighboring MVs, the hit-ratio in the previous CUs, one of symmetric bilateral matching or asymmetric bilateral matching may be be implicitly selected.

In one example, if POC_CURis greater than POC₀and POC_CURis greater than POC₁ , then the symmetric bilateral matching is performed. Otherwise, the asymmetric bilateral matching is performed.

In one example, if the current motion vector predictor has both forward MV0 and backward MV1 (i.e., if the current motion vector predictor is bi-prediction), then the asymmetric bilateral matching is performed. Otherwise (i.e., if the current motion vector predictor is uni-prediction), the symmetric bilateral matching is performed.

As described above, in JEM, in cases where a candidate neighboring video block is predicted using template matching is applied independently to a first reference picture and a second reference picture and one of a uni-prediction or a bi-prediction is selected. That is, in JEM, the distortions (e.g. Sum of absolute difference) of the first/second uni-prediction are measured between the first/second uni-prediction and the reconstructed image data in neighboring blocks. The distortions of each uni-prediction are compared with the bi-directional distortion, which is measured between the first prediction and the second prediction using the bilateral matching cost. Among two uni-prediction distortions and one bi-prediction distortion, the best one having the minimum distortion is selected. In one example, according to the techniques herein, inter prediction processing unit 214 may be configured to construct a new predicted image data using two MVs derived from uni-direction template matching. That is, for example, sample values in a template shape corresponding the best match in a forward direction and sample values in a template shape corresponding the best match in a backward direction may be averaged to construct a new prediction image data. The distortion between the new predicted image and the reconstructed image in neighbor CU is measured. If the measured distortion is smaller than the distortion of Uni-prediction, Bi-prediction is selected.

As described above, in order to perform template matching and bilateral matching a search may be performed according to a search pattern initiated with respect to an origin point. According to the techniques herein, in template matching and bilateral matching a search may be performed according to a hierarchical search. For example, inter prediction processing unit 214 may be configured to perform a hierarchical search according to defined shapes (e.g., rectangle, diamond, cross) and step sizes (i.e., offset values), where a shape is specified at each level of a hierarchy and the step size is defined at each level of the hierarchy, where the step size is reduced at each level of the hierarchy. For example, FIG. 13 illustrates an example of using a hierarchical search for template matching. Referring to FIG. 13, a top level of a hierarchy may be associated with a square shape having steps {S₀, S₁, S₂, S₃}. That is, each of {S₀, S₁, S₂, S₃} may correspond to an offset (e.g., (|8|,|8|) indicating a position of a set of samples to compare to a template. Further, at each of steps {S₀, S₁, S₂, S₃} another shape and set steps may be specified. For example, as illustrated in FIG. 13, for S₀a cross shape having steps {S₀₀, S₀₁, S₀₂, S₀₃} is specified. In this manner, in the example illustrated in FIG. 13, the value of the step size may be reduced at each level of the hierarchical search.

In one example, performing a search may include performing a search at each level for all origin points. For example, for each motion vector candidate, determine the best match at level 0. In one example, the best match of all of the motion vector candidates becomes a new origin for the next round and the round is iteratively repeated until the new best match is identical to the previous best one (i.e., the search converged). If the search does not converge and the current number of rounds exceeds a predefined count, the level is completed and the next level search is initiated. The predefined count may be unique to each level. For example, the predefined count may be large (e.g., 4, 8, or even larger) for the first level of the hierarchy, and equal to 1 for the following levels.

In one example, an extended search pattern can be used. The extended search pattern may have multiple layers of candidates. For example, it may have two layers. A first layer has candidates close to the origin, while a second layer has candidates further away from the origin. If the best candidate is selected among one of candidates in the first layer, the current level is completed and the search process moves on to the next level in the hierarchy. Otherwise, the next round is launched with the current best candidate as a new origin. Thus a next round is initiated only when one of candidates from the second layer is selected as the best.

In one example, the search patterns and range can be adaptively changed according to one or more of the following factors: Vertical or Horizontal direction of MV, CU/PU/TU size, POC distance between the current slice and the reference slice, adjacent MVs or MVD values, variance of adjacent MVs and MVD values, temporal depth in a hierarchical B structure, QT, BT, or QTBT combined depth, slice type (P or B), inter-prediction type (forward, backward, bi-prediction), quantization parameter, CU-level or PU-level coding, and/or accumulated statistics.

In one example, the search range for an motion vector origin in asymmetric bilateral matching may be determined as an intersecting region between a forward search range and a backward search range. For example, referring to FIG. 15A, a backward motion vector origin, MV0 (mvx0, mvy0), may form the center of an initial search range having dimensions top0, bottom0, left0, and right0, and a forward motion vector origin, MV1 (mvx1, mvy1), may form the center of an initial search range having dimensions top1, bottom1, left1, and right1. A restricted search range for MV0 may have dimensions top0’, bottom0’, left0’, and right0’ and a restricted search range for MV1 may have dimensions top1’, bottom1’, left1’, and right1’. Referring to FIG. 15B, FIG, 15B illustrates an example where top0’, bottom0’, left0’, and right0’ and top1’, bottom1’, left1’, and right1’ are determined as the overlapping regions of the initial search range for MV0 and MV1. In one example, top0’, bottom0’, left0’, and right0’ may be determined by excluding region of top0, bottom0, left0, and right0 not encapsulated by the region defined by top1, bottom1, left1, and right1. In one example, top1’, bottom1’, left1’, and right1’ may be determined by excluding the region of top1, bottom1, left1, and right1 not encapsulated by the region defined by top0’, bottom0’, left0’, and right0’. In one example, top1’, bottom1’, left1’, and right1’ may be determined by centering top0’, bottom0’, left0’, and right0’ at MV1.

In one example, with respect to MV0 and MV1 and the respective initial search ranges the following initial values may be defined:

In one example, a variable a may be calculated as:

In one example, when a is greater than 0, topDist0; bottomDist0; leftDist0; rightDist0; topDist1; bottomDist1; leftDist1; and rightDist1 may be updated as follows:

and when a is less than 0, topDist0; bottomDist0; leftDist0; rightDist0; topDist1; bottomDist1; leftDist1; and rightDist1 may be updated as follows:

Thus, the restricted search range for MV0 and MV1 may be determined according to the following:

It should be noted that in some cases, for parallel processing, accessing spatially adjacent motion vectors across a boundary may be prohibited. In one example, for any CU boundary access to neighboring motion vectors for determining a motion vector by a decoder-side motion derivation process (e.g. bilateral matching) may be allowed/disallowed. In one example, a flag may indicate with access is allowed for each boundary or CU. In one example, all CU boundaries of a picture/slice/sequence can allow/disallow the access by explicit signaling.

In one example, motion vector determination by a decoder-side motion vector derivation process may not be used for determining a motion vector for any other CU in the current frame, however, the decoder-side motion vector derivation process may be used for determining the motion vectors for a future frame. Such a restriction, anticipates that a decoder may need to decode a motion vector well in advance of decoding the previous CU, in order to optimize memory access. Such a restriction may be signaled using a flag in a parameter set, a slice header, a supplemental enhancement message, and/or at the CU-level. In one example, when a motion vector is determined by a decoder-side MV derivation process (including e.g., bilateral matching) in the current CU is not used for predicting/coding motion vectors of the following CUs in the current frame, the motion vector predictor derived by Merge/AMVP/Affine modes in the current CU may be used for predicting/coding motion vectors of the following CUs in the current frame, instead.

It should be noted that although the techniques herein are described in the examples above with respect to CUs and CU sub-blocks, the techniques described herein may be generally applicable to video blocks and sub-divisions thereof.

Referring again to FIG. 10, as illustrated in FIG. 10, inter prediction processing unit 214 may receive reconstructed video block via filter unit 216, which may be part of an in-loop filtering process. Filter unit 216 may be configured to perform deblocking and/or Sample Adaptive Offset (SAO) filtering. Deblocking refers to the process of smoothing the boundaries of reconstructed video blocks (e.g., make boundaries less perceptible to a viewer). SAO filtering is a non-linear amplitude mapping that may be used to improve reconstruction by adding an offset to reconstructed video data. Entropy encoding unit 218 receives quantized transform coefficients and predictive syntax data (i.e., intra prediction data, motion prediction data, QP data, etc.). Entropy encoding unit 218 may be configured to perform entropy encoding according to one or more of the techniques described herein. Entropy encoding unit 218 may be configured to output a compliant bitstream, i.e., a bitstream that a video decoder can receive and reproduce video data therefrom. In this manner video encoder 200 represents an example of a device configured to for each location in a search pattern about one or more origin points, determine a cost for a pair of corresponding blocks, wherein the cost is determined based on the absolute sum of differences of the corresponding blocks and a side distortion cost and select a motion vector predictor as the motion vector associated with the lowest cost.

FIG. 14 is a block diagram illustrating an example of a video decoder that may be configured to decode video data according to one or more techniques of this disclosure. In one example, video decoder 300 may be configured to reconstruct video data based on one or more of the techniques described above. That is, video decoder 300 may operate in a reciprocal manner to video encoder 200 described above. Video decoder 300 may be configured to perform intra prediction decoding and inter prediction decoding and, as such, may be referred to as a hybrid decoder. In the example illustrated in FIG. 14 video decoder 300 includes an entropy decoding unit 302, inverse quantization unit 304, inverse transformation processing unit 306, intra prediction processing unit 308, inter prediction processing unit 310, summer 312, filter unit 314, and reference buffer 316. Video decoder 300 may be configured to decode video data in a manner consistent with a video encoding system, which may implement one or more aspects of a video coding standard. It should be noted that although example video decoder 300 is illustrated as having distinct functional blocks, such an illustration is for descriptive purposes and does not limit video decoder 300 and/or sub-components thereof to a particular hardware or software architecture. Functions of video decoder 300 may be realized using any combination of hardware, firmware, and/or software implementations.

As illustrated in FIG. 14, entropy decoding unit 302 receives an entropy encoded bitstream. Entropy decoding unit 302 may be configured to decode quantized syntax elements and quantized coefficients from the bitstream according to a process reciprocal to an entropy encoding process. Entropy decoding unit 302 may be configured to perform entropy decoding according any of the entropy coding techniques described above. Entropy decoding unit 302 may parse an encoded bitstream in a manner consistent with a video coding standard. Video decoder 300 may be configured to parse an encoded bitstream where the encoded bitstream is generated based on the techniques described above. Inverse quantization unit 304 receives quantized transform coefficients (i.e., level values) and quantization parameter data from entropy decoding unit 302. Quantization parameter data may include any and all combinations of delta QP values and/or quantization group size values and the like described above. Video decoder 300 and/or inverse quantization unit 304 may be configured to determine QP values used for inverse quantization based on values signaled by a video encoder and/or through video properties and/or coding parameters. That is, inverse quantization unit 304 may operate in a reciprocal manner to coefficient quantization unit 206 described above. Inverse quantization unit 304 may be configured to apply an inverse quantization. Inverse transform processing unit 306 may be configured to perform an inverse transformation to generate reconstructed residual data. The techniques respectively performed by inverse quantization unit 304 and inverse transform processing unit 306 may be similar to techniques performed by inverse quantization/transform processing unit 208 described above. Inverse transform processing unit 306 may be configured to apply an inverse DCT, an inverse DST, an inverse integer transform, Non-Separable Secondary Transform (NSST), or a conceptually similar inverse transform processes to the transform coefficients in order to produce residual blocks in the pixel domain. Further, as described above, whether a particular transform (or type of particular transform) is performed may be dependent on an intra prediction mode. As illustrated in FIG. 14, reconstructed residual data may be provided to summer 312. Summer 312 may add reconstructed residual data to a predictive video block and generate reconstructed video data.

As describe above, a predictive video block may be determined according to a predictive video technique (i.e., intra prediction and inter frame prediction). Intra prediction processing unit 308 may be configured to receive intra prediction syntax elements and retrieve a predictive video block from reference buffer 316. Reference buffer 316 may include a memory device configured to store one or more frames of video data. Intra prediction syntax elements may identify an intra prediction mode, such as the intra prediction modes described above. In one example, intra prediction processing unit 308 may reconstruct a video block using according to one or more of the intra prediction coding techniques described herein. Inter prediction processing unit 310 may receive inter prediction syntax elements and generate motion vectors to identify a prediction block in one or more reference frames stored in reference buffer 316. Inter prediction processing unit 310 may produce motion compensated blocks, possibly performing interpolation based on interpolation filters. Identifiers for interpolation filters to be used for motion estimation with sub-pixel precision may be included in the syntax elements. Inter prediction processing unit 310 may use interpolation filters to calculate interpolated values for sub-integer pixels of a reference block.

As described above, video decoder 300 may parse an encoded bitstream where the encoded bitstream is generated based on the techniques described above and as described above video encoder 200 may generate a bitstream according to the motion vector prediction techniques described above. In this manner video decoder 300 represents an example of a device configured to for each location in a search pattern about one or more origin points, determine a cost for a pair of corresponding blocks, wherein the cost is determined based on the absolute sum of differences of the corresponding blocks and a side distortion cost and select a motion vector predictor as the motion vector associated with the lowest cost.

Referring again to FIG. 14, filter unit 314 may be configured to perform filtering on reconstructed video data. For example, filter unit 314 may be configured to perform deblocking and/or SAO filtering, as described above with respect to filter unit 216. Further, it should be noted that in some examples, filter unit 314 may be configured to perform proprietary discretionary filter (e.g., visual enhancements). As illustrated in FIG. 14 a reconstructed video block may be output by video decoder 300.

In one or more examples, the functions described may be implemented in hardware, software, firmware, or any combination thereof. If implemented in software, the functions may be stored on or transmitted over as one or more instructions or code on a computer-readable medium and executed by a hardware-based processing unit. Computer-readable media may include computer-readable storage media, which corresponds to a tangible medium such as data storage media, or communication media including any medium that facilitates transfer of a computer program from one place to another, e.g., according to a communication protocol. In this manner, computer-readable media generally may correspond to (1) tangible computer-readable storage media which is non-transitory or (2) a communication medium such as a signal or carrier wave. Data storage media may be any available media that can be accessed by one or more computers or one or more processors to retrieve instructions, code and/or data structures for implementation of the techniques described in this disclosure. A computer program product may include a computer-readable medium.

By way of example, and not limitation, such computer-readable storage media can comprise RAM, ROM, EEPROM, CD-ROM or other optical disk storage, magnetic disk storage, or other magnetic storage devices, flash memory, or any other medium that can be used to store desired program code in the form of instructions or data structures and that can be accessed by a computer. Also, any connection is properly termed a computer-readable medium. For example, if instructions are transmitted from a website, server, or other remote source using a coaxial cable, fiber optic cable, twisted pair, digital subscriber line (DSL), or wireless technologies such as infrared, radio, and microwave, then the coaxial cable, fiber optic cable, twisted pair, DSL, or wireless technologies such as infrared, radio, and microwave are included in the definition of medium. It should be understood, however, that computer-readable storage media and data storage media do not include connections, carrier waves, signals, or other transitory media, but are instead directed to non-transitory, tangible storage media. Disk and disc, as used herein, includes compact disc (CD), laser disc, optical disc, digital versatile disc (DVD), floppy disk and Blu-ray disc where disks usually reproduce data magnetically, while discs reproduce data optically with lasers. Combinations of the above should also be included within the scope of computer-readable media.

Instructions may be executed by one or more processors, such as one or more digital signal processors (DSPs), general purpose microprocessors, application specific integrated circuits (ASICs), field programmable logic arrays (FPGAs), or other equivalent integrated or discrete logic circuitry. Accordingly, the term “processor,” as used herein may refer to any of the foregoing structure or any other structure suitable for implementation of the techniques described herein. In addition, in some aspects, the functionality described herein may be provided within dedicated hardware and/or software modules configured for encoding and decoding, or incorporated in a combined codec. Also, the techniques could be fully implemented in one or more circuits or logic elements.

The techniques of this disclosure may be implemented in a wide variety of devices or apparatuses, including a wireless handset, an integrated circuit (IC) or a set of ICs (e.g., a chip set). Various components, modules, or units are described in this disclosure to emphasize functional aspects of devices configured to perform the disclosed techniques, but do not necessarily require realization by different hardware units. Rather, as described above, various units may be combined in a codec hardware unit or provided by a collection of interoperative hardware units, including one or more processors as described above, in conjunction with suitable software and/or firmware.

Moreover, each functional block or various features of the base station device and the terminal device used in each of the aforementioned embodiments may be implemented or executed by a circuitry, which is typically an integrated circuit or a plurality of integrated circuits. The circuitry designed to execute the functions described in the present specification may comprise a general-purpose processor, a digital signal processor (DSP), an application specific or general application integrated circuit (ASIC), a field programmable gate array (FPGA), or other programmable logic devices, discrete gates or transistor logic, or a discrete hardware component, or a combination thereof. The general-purpose processor may be a microprocessor, or alternatively, the processor may be a conventional processor, a controller, a microcontroller or a state machine. The general-purpose processor or each circuit described above may be configured by a digital circuit or may be configured by an analogue circuit. Further, when a technology of making into an integrated circuit superseding integrated circuits at the present time appears due to advancement of a semiconductor technology, the integrated circuit by this technology is also able to be used.

Various examples have been described. These and other examples are within the scope of the following claims.

<Cross Reference>
This Nonprovisional application claims priority under 35 U.S.C. § 119 on provisional Application No. 62/651,056 on March 30, 2018, No. 62/679,673 on June 1, 2018, the entire contents of which are hereby incorporated by reference.

Claims

A method of determining a motion vector predictor for coding video data, the method comprising:
for each location in a search pattern about one or more origin points, determining a cost for a pair of corresponding blocks, wherein the cost is determined based on the absolute sum of differences of the corresponding blocks and a side distortion cost; and
selecting a motion vector predictor as the motion vector associated with the lowest cost.
The method of claim 1, wherein a side distortion cost includes a side template match distortion cost.
The method of claim 1, wherein a side distortion cost includes a side continuity distortion cost.
The method of any of claims 1-3, wherein the cost is determined based on a weighted sum of the absolute sum of differences of the corresponding blocks and a side distortion cost.
The method of any of claims 1-4 wherein the search pattern includes a hierarchical search pattern.
The method of any of claims 1-4 wherein the pair of corresponding blocks is determined using a forward motion vector and a backward motion vector which are asymmetric.
A device comprising one or more processors configured to perform any and all combinations of the steps of claims 1-6.
The device of claim 7, wherein the device includes a video encoder.
The device of claim 7, wherein the device includes a video decoder.
A system comprising:
the device of claim 8; and
the device of claim 9.
An apparatus for coding video data, the apparatus comprising means for performing any and all combinations of the steps of claims 1-6.
A non-transitory computer-readable storage medium comprising instructions stored thereon that, when executed, cause one or more processors of a device for coding video data to perform any and all combinations of the steps of claims 1-6.