Back to Search Document details
13th Meeting: Marrakech, January 2019 2019-01-17 10:39
CE4-related: Pairwise Average Candidate Reduction
Abstract
In this contribution, it is proposed to reduce the number of pairwise average candidates added in the merge list. There are up to six pairwise average candidates defined during the generation of merge candidate list. However as shown in this proposal, it is never possible for some of the candidates to be used in the list generation. Also, other candidates that are used in the list generation, don’t improve the performance but add additional complexity for generation of the list. This contribution proposes several options to reduce the number of pairwise average candidates. The performance change by just using only one pairwise average candidate when tested on top of VTM3.0 under RA conditions is 0.01% for Y, 0.03% for U and 0.02% for V with no change in runtime.
JVET-M0193 CE4-related: Pairwise Average Candidate Reduction [A. Tamse, M. W. Park, K. Choi (Samsung)]

Notes from BoG report review: 0.01%/0.00% loss in RA/LB, reduce pairwise candidate from 6 to 1

JVET-M0193 CE4-related: Pairwise Average Candidate Reduction [A. Tamse, M. W. Park, K. Choi (Samsung)]

Notes from BoG report review: 0.01%/0.00% loss in RA/LB, reduce pairwise candidate from 6 to 1

PATENTS:
US11202065B2 0.34 2021-12-14 US12323617B2 0.32 2025-06-03 US20250030839A1 0.30 2025-01-23 CN114503563B 0.28 2025-06-13 JP7367018B2 0.26 2023-10-23 CN113170184B 0.24 2024-12-06 CN113261290B 0.22 2024-03-12 CN113455003B 0.20 2023-09-29
Decisions
JVET-M0193 CE4-related: Pairwise Average Candidate Reduction
Notes from BoG report review: 0.01%/0.00% loss in RA/LB, reduce pairwise candidate from 6 to 1
Citation