JVET-Q0819 AHG2: Editorial input of integrated text of PH changes [Y.-K. Wang (Bytedance)] [late]
This contribution was discussed Thursday 14 January in Track A at 0950 (chaired by GJS).
This input contribution is an editing input to provide text for the decisions that have been recorded regarding a number of PH changes.
It was noted that there is no byte alignment between the PH-in-SH data and the SH data.
Some issues were noted in the review and are to be corrected in a revision (e.g., inference of ph_intra_slices_allowed_flag equal to 1 if ph_inter_slices_allowed_flag is equal to 0, dbf_info_in_ph_flag shouldn’t be in the PH, conditioning slice_qp_delta on qp_delta_info_in_ph_flag, do not include slice_cu_qp_delta_subdiv at the slice level, mvd_l1_zero_flag should be sent unconditionally in the SH for inter slices).
It was noted that, unlike some other cases, gdr_pic_flag is not defined with “one way” semantics. This aspect was not modified.
The weighted prediction part had open issues at the time this was reviewed.
Decision (Ed.): Adopt as expressing the decisions recorded elsewhere in these notes.
There was discussion of whether the software would have one large macro switch or there would be a large number of smaller incremental changes as per the usual software guideline. Since this has a number of interacting changes, it was agreed that there should be pre-coordination of their integration and does not need to be separate macros for each aspect. J. Samuelsson volunteered to coordinate that.
Mixed NAL unit types within a coded picture (7)
Contributions in this category were discussed in Track A Friday 10 January 2020 at 1500 (chaired by GJS).