Back to Search Document details
26th Meeting: by teleconference, April 2022 2022-04-14 06:40
EE2-2.5: Adaptive reordering of merge candidates with refined motion
Authors: Yifan Wang ByteDance Kai Zhang ByteDance N. Zhang ByteDance Zhipin Deng ByteDance Li Zhang ByteDance Yang Wang
Abstract
This contribution reports test results for EE2-2.5 on adaptive reordering of merge candidates (ARMC) with refined motion. In ARMC with refined motion, each merge candidate in the merge candidate list is refined using TM/multi-pass DMVR first and the refined motion will be used in ARMC to reorder the merge candidate list. When multi-pass DMVR is used to derive the refined motion, only the first pass (i.e., PU level) of multi-pass DMVR is applied in reordering. ARMC with refined motion is tested in Test #2.5a, while a simplified version is tested in Test #2.5b. On top of ECM-4.0, simulation results of EE2-2.5 are reported as below:
References:
JVET-Y0151 JVET-Y2017 JVET-Y2024
Cited:
JVET-Z0145 JVET-Z0163 X
PATENTS:
20240155149 1.00 WO/2023/284817 0.95 20240275943 0.90 20240275984 0.85 20240267533 0.80 WO/2022/253319 0.75 WO/2022/253318 0.70 WO/2022/253320 0.65
ARCHIVE v1 :
JVET-Z0058-Test1.xlsm JVET-Z0058-Test2.xlsm JVET-Z0058.docx
Citation