JVET-R0097 AHG9: Transform and transform-skip related HLS clean-up [M. G. Sarwer, Y. Ye, J. Luo, J. Chen (Alibaba)]
Aspect 1 of this was initially discussed in Track B on Saturday 18 April, and the notes were then moved here to be better categorized.
This contribution proposes two aspects to clean up the signalling of sps_max_luma_transform_size_64_flag and slice_ts_residual_coding_disabled_flag.
- In aspect 1, if the luma CTB size is not larger than 32, sps_max_luma_transform_size_64_flag is not signalled and inferred to be 0.
- In aspect 2, two methods are proposed to reduce the signalling overhead of slice_ts_residual_coding_disabled_flag in slice header.
- Method 1: an SPS level flag is introduced to indicate the presence of slice_ts_residual_coding_disabled_flag in the bitstream. The proposed SPS level flag is signalled if the sps_transform_skip_enabled_flag is equal to 1.
- Method 2: if the sps_transform_skip_enabled_flag is equal to 0, slice_ts_residual_coding_disabled_flag is not signalled and inferred to be 0.
The notes were moved here from the AHG cat. 2 discussions and recommendations of track B
See also the notes under JVET-R0049 for aspect 2.
Aspect 1 was presented Sat 18 Apr. It is proposed to replace the existing bitstream constraint by a syntax constraint. This was asserted to be cleaner from LL perspective and is suggested to be adopted from the side of track B – was further discussed in Sunday plenary and agreed. Decision (cleanup): Adopt JVET-R0097 aspect 1.
It is mentioned that also a ticket #1024 was issued that pointed out a conformance stream violating the 32×32 bitstream constraint. This was a fortuitous ticket number for an issue relating to a 32×32 block size issue.