JVET-O1021 Crosscheck of JVET-O0186 (AHG15: Signalling of chroma Qp mapping table) [K. Choi (Samsung)]

General consideration of chroma QP mapping

Discussion time Thursday 4 July 2110-2230 (GJS)

Issues:

• Separate mapping tables for Cb and Cr – yes, should support this (also supporting a common table)
• SPS, PPS, or APS – SPS
• Default(s) – no default
• Additional offsets prior to or after the mapping table (currently we have PPS and slice level offsets before the mapping) – PPS and slice level, offsets applied after the mapping table (deblocking continues to use only the picture level offset)
• Joint chroma coding handling – use a third table if not using a common table, and support a third offset at PPS and slice level (similar to what we’re doing now at the PPS and slice level)
• Additional issues:
  • Degree of generality needed for the mapping table (slope constraints other than non-negative, and starting point constraints) – it was commented that we seem to have some use for a general non-negative slope, where we want slope of 1 at very low QP, a somewhat higher QP for chroma in some middle range, but then a somewhat lower QP for chroma at very high QP values – we prefer to have a general syntax, O0562 actually supports negative slopes, O0650 supports arbitrary non-negative slopes, so it was suggested to consider that syntax as a candidate for adoption; see further notes after offline study.
  • Syntax & semantics details (amount of bits, number of syntax elements, text description)
  • What to do in CTC

This was further discussed Thursday 11 July at 1245 (GJS).

Text was presented as an additional document in the zip archive for JVET-O0650. The tables are also sent for 4:4:4 and 4:2:0. Decision: Adopt as in the new attachment.

For CTC, it was initially agreed to use the new syntax to achieve the chroma QP settings proposed in JVET-O0186 for SDR, N0221 for PQ, O0433 for HLG. For the joint table, use the average of the Cb and Cr tables. Below QP = 16, chroma and luma QP should be equal for CEs using low QP.

Decision (CTC): This was further discussed in the closing plenary on Friday 12 July. It was suggested to keep the anchor QP settings for PQ and HLG as in CTC of prior meeting cycle. For SDR, see above. Below QP = 16, chroma and luma QP should be equal for CEs using low QP.

Further study of CTC was encouraged.

See the notes below for JVET-O1126 and JVET-O1166, which were also briefly discussed on Thursday 11 July.