US 11,902,535 B2
Prediction precision improvements in video coding
Kai Zhang, San Diego, CA (US); Li Zhang, San Diego, CA (US); Hongbin Liu, Beijing (CN); and Yue Wang, Beijing (CN)
Assigned to BEIJING BYTEDANCE NETWORK TECHNOLOGY CO., LTD, Beijing (CN); and BYTEDANCE INC., Los Angeles, CA (US)
Filed by Beijing Bytedance Network Technology Co., Ltd., Beijing (CN); and Bytedance Inc., Los Angeles, CA (US)
Filed on Oct. 19, 2021, as Appl. No. 17/505,227.
Application 17/505,227 is a continuation of application No. 17/167,169, filed on Feb. 4, 2021, abandoned.
Application 17/167,169 is a continuation of application No. PCT/CN2019/115722, filed on Nov. 5, 2019.
Claims priority of application No. PCT/CN2018/113928 (WO), filed on Nov. 5, 2018.
Prior Publication US 2022/0038714 A1, Feb. 3, 2022
Int. Cl. H04N 19/159 (2014.01); H04N 19/172 (2014.01); H04N 19/176 (2014.01); H04N 19/96 (2014.01); H04N 19/117 (2014.01); H04N 19/132 (2014.01); H04N 19/186 (2014.01); H04N 19/105 (2014.01); H04N 19/513 (2014.01)
CPC H04N 19/159 (2014.11) [H04N 19/105 (2014.11); H04N 19/117 (2014.11); H04N 19/132 (2014.11); H04N 19/172 (2014.11); H04N 19/176 (2014.11); H04N 19/186 (2014.11); H04N 19/513 (2014.11); H04N 19/96 (2014.11)] 16 Claims
OG exemplary drawing
 
1. A method of processing video data, comprising:
determining, for a conversion between a current block of a video and a bitstream of the video, two corresponding regions for a sub-block of the current block when a first prediction mode is applied to the current block, wherein the two corresponding regions are in two reference pictures of the current block respectively, and wherein the first prediction mode is an optical flow-based inter prediction mode;
deriving a sum of absolute difference (SAD) based on partial samples of the two corresponding regions, wherein positions of the partial samples of the two corresponding regions are predetermined,
determining the first prediction mode is skipped for the sub-block in response to the SAD being smaller than a first threshold,
deriving, when the first prediction mode is not skipped for the sub-block,
spatial gradient values in different directions based on samples of the corresponding regions and an arithmetic shifting operation with a shift value S which is a fixed integer and is not equal to 4,
generating prediction samples for the current block based on the spatial gradient values and an interpolation filtering operation, and
performing the conversion based on the prediction samples,
wherein the spatial gradient values are derived based on samples from a same reference picture,
wherein the shift value S is independent of a sample bit-depth of the current block,
wherein a size of the corresponding region is M×N, and a first number of samples required to derive the spatial gradient values is (M+G)×(N+G),
wherein an interpolation filter comprising L taps is used in the interpolation filtering operation for a luma component of the sub-block,
wherein a number of samples required by the interpolation filtering operation is (M+k+L−1)×(N+k+L−1), wherein k is an integer less than G,
wherein whether the first prediction mode is applied to the current block is determined based on τ0 and τ1, wherein τ0 is a difference between a picture order count of the current picture and a picture order count of a first reference picture of the two reference pictures (τ0=POC(current)−POC(R0)), and wherein τ1 is a difference between a picture order count of a second reference picture of the two reference pictures and the picture order count of the current picture (τ1=POC(R1)−POC(current)), and
wherein the first prediction mode is not applied to the current block in response to abs(abs(τ0)−abs(τ1))>=T wherein T is an integer.