Back to Search Document details
13th Meeting: Marrakech, January 2019 2019-01-07 15:28
CE4-related: CE4-related: Quadtree-based Merge Estimation Region for VVC
Authors: T.L. Fu Peking Univ. Xiaozhen Zheng DJI S.S Wang S.W. Ma
Abstract
In this contribution, a method for parallel merge estimation of neighboring blocks is proposed, which can handle quadtree, binary tree and ternary tree partitions. In principal, the proposed method sets the leaf node of quad-tree as the size of merge estimation region (MER), which means candidate blocks in the same leaf node of quad-tree is not included in the merge candidate list construction. Simulation results for N equal to 8, 16, 32 and 64 are presented in the contribution document.
Meeting Notes
JVET-M0350 CE4-related: Quadtree-based Merge Estimation Region for VVC [T. L. Fu (Peking Univ.), X. Zheng (DJI), S. S Wang, S. W. Ma (Peking Univ.)]

Notes from BoG report review: 0.49% (94%, 90%)/0.49% (93%, 91%)

In the follow-up discussion in Track B, it was asked whether there would be much difference if it was implemented as encoder-only approach? It is probably good for some encoder runtime reduction, but has also quite some loss, so we would likely not adopt it or put it in CTC.

JVET-M0350 CE4-related: Quadtree-based Merge Estimation Region for VVC [T. L. Fu (Peking Univ.), X. Zheng (DJI), S. S Wang, S. W. Ma (Peking Univ.)]

Notes from BoG report review: 0.49% (94%, 90%)/0.49% (93%, 91%)

In the follow-up discussion in Track B, it was asked whether there would be much difference if it was implemented as encoder-only approach? It is probably good for some encoder runtime reduction, but has also quite some loss, so we would likely not adopt it or put it in CTC.

References:
JVET-L0216
Cited:
JVET-M0024 JVET-M0583 X JVET-M0851
ARCHIVE v2 :
JVET-M0350-MER16x16.xlsm JVET-M0350-MER32x32.xlsm JVET-M0350-MER64x64.xlsm JVET-M0350-MER8x8.xlsm JVET-M0350_v2.docx
Decisions
Related to JVET-M0281
adopted
In the follow-up discussion in Track B, it was asked whether there would be much difference if it was implemented as encoder-only approach? It is probably good for some encoder runtime reduction, but has also quite some loss, so we would likely not adopt it or put it in CTC.
Citation