JVET-P1001 Non-CE5: Chroma QP derivation fix for deblocking filter (Combination of JVET-P0105 and JVET-P0539) [A. M. Kotra, S. Esenlik, B. Wang, H. Gao, E. Alshina (Huawei)] [late]
It is claimed that there is a problem that table for Joint CbCr mode not used in the deblocking although used in transform coefficient scaling.
Two solutions are proposed. Solution 1 uses only the picture level QP offset.
Solution 2 uses slice level and block level QP offsets.
The proponent thinks that Solution 1 is more in the spirit of the VTM behavior, while the Solution 2 better takes into account the transform coefficient scaling.
There are two aspects of this proposal: solution 1 fixes the chroma mapping problem, whereas solution 2 also fixes the aspect of QP offsets.
It was noted that there is a note in the specification that suggests that this problem was not fixed (HEVC), however, none of the participants remember why.
Some participants mentioned that fundamentally you would like to solve this problem. Some participants expressed concerns regarding complexity.
BoG recommendation: Adopt solution 2 (if there are no complexity concerns raised by the time of review in the track).