JVET-AD0236 Non-EE2: On Classification of In-Loop Filters [W. Yin, K. Zhang, L. Zhang (Bytedance)]

In this contribution, three additional rules for classification of in-loop filters are proposed, including deblocking filter-boundary strength (DBF-BS) based classification for ALF, DBF-BS based classification for SAO and variance based classification for biliteral filter (BF).

On top of ECM-8.0, simulation results of the proposed methods are reported as below:

AI: -0.01%, -0.03%, 0.04%, 100%, 100%.

RA: %, %, %, %, %.

LB: %, %, %, %, %.

For AI, gain is not interesting

Expected in RA: 0.1% luma, most gain comes from class C

What are the contributions of the three elements which could be implemented completely independent?

No support by other experts for investigation in EE. Further study was recommended.