JVET-R0049 AHG9: HLS on disabling TSRC [S.-T. Hsiang, C.-W. Hsu, Z.-Y. Lin, T.-D. Chuang, C.-Y. Chen, Y.-W. Huang, S.-M. Lei (MediaTek)]

This contribution and related contributions were initially reviewed in the Category 2 AHG pre-meeting as reported in JVET-R0340, then in a JVET plenary on Sunday 19 April. The notes were moved here from section 5.3.1 since this is an HLS topic. (There are other items in section 5.3.1 that are related, as the matter involves both high-level and low-level issues.)

In VVC Draft 8, extra bit costs are consumed for signalling the new syntax element slice_ts_residual_coding_disabled_flag for each of the coded slices. This contribution proposes three methods for high-level syntax modifications related to disabling transform skip residual coding, as follows:

• In Method 1, a new syntax element sps_ts_residual_coding_disabled_slice_present_flag is added to the sequence parameter set (SPS) to specify whether slice_ts_residual_coding_disabled_flag is present in the slice header.
• In Method 2, two new syntax elements sps_ts_residual_coding_disabled_slice_present_flag and sps_ts_residual_coding_disabled_slice_default_flag are added to the SPS. When slice_ts_residual_coding_disabled_flag is not present, it is inferred to be equal to the value of sps_ts_residual_coding_disabled_slice_default_flag.
• In Method 3, signalling slice_ts_residual_coding_disabled_flag in the slice header is conditioned on sps_transform_skip_enabled_flag equal to 1 without adding any new syntax element.

Method 1 same is in JVET-R0097.

Method 3 same as in JVET-R0097, JVET-R0068 (item 8) plus inference to 1 for slice_ts_residual_coding_disabled_flag, JVET-R0142, JVET-R0317 without the PPS flag, JVET-R0153 aspect 1, JVET-R0182 with inverse semantics.

The motivation behind method 1 is to not signal slice_ts_residual_coding_disabled_flag for lossy coding cases, which are considered to be the main application.

It was commented that at the last meeting, slice_ts_residual_coding_disabled_flag was adopted as a slice level flag instead of PPS or SPS flag. This would not prevent syntax as proposed here to gate its presence.

A participant questions whether the additional control syntax to save the signalling of slice_ts_residual_coding_disabled_flag is really needed. Functionality is not affected by any of the proposed methods.

It was agreed that it is desirable to not always send slice_ts_residual_coding_disabled_flag since it is expected to only be used for lossless coding scenarios.

It was further agreed that it makes sense to condition slice_ts_residual_coding_disabled_flag on sps_transform_skip_enabled_flag (Method 3).

Method 3 as discussed (in combination with JVET-R0271):

SPS: sps_transform_skip_enabled_flag

SH: if( sps_transform_skip_enabled_flag && !( sh_SDH  | |  sh_DQ) )

slice_ts_residual_coding_disabled_flag

Method 1 as discussed (in combination with JVET-R0271):

SPS: sps_transform_skip_enabled_flag

SPS: if( sps_transform_skip_enabled_flag )

sps_ts_residual_coding_disabled_flag

SH: if( !sps_ts_residual_coding_disabled_flag && !( sh_SDH  | |  sh_DQ ) )

slice_ts_residual_coding_disabled_flag

Method 1 would allow an encoder to disable TSRC completely (e.g., if it only want to support one kind of residual coding). Method 3 would not allow this unless it also enables SDH or DQ.

We don’t want the combination of SDH/DQ enabled with TS enabled and TSRC disabled.

Variant 2 of JVET-R0271 was also discussed.

It was requested for offline work with proponents of JVET-R0049 and JVET-R0271 and document submission for a clear similar syntax description and truth table analysis of combinations of variants 2 and 3 of JVET-R0271 with methods 1 and 3 of JVET-R0049. See documents JVET-R0483, JVET-R0485, JVET-R0486 which were submitted later.

uses of interest: lossy and lossless regions, with

• lossy & lossless screen content regions
• lossy & lossless camera-captured content

The plenary session of 19 April stopped here at 1500.