JVET-R0388 AHG9: Cleanups on deblocking signalling [Z. Deng, Y.-K. Wang, L. Zhang, K. Zhang, J. Xu (Bytedance)]
This contribution was discussed in JVET on Wednesday 22 April at about 0845 UTC (chaired by GJS).
After the categroy 1 AHG pre-meeting review of the contributions on deblocking signalling control, the proponents further studied the topic taking into account the AHG recommendation. It is asserted that 1) there are still some issues with the semantics that should be fixed, 2) there is a bug in the inference of the slice_deblocking_filter_override_flag, and 3) the feature of indicating an overriding operation in a picture header or a slice header and then immedaitely send the next bit in the same header to indicate a change of mind is a bit too weird and should be removed.
To address the above issues, this contribution proposes the following changes:
- Change the semantics of the deblocking signalling control syntax elements as editorial improvement.
- Infer slice_deblocking_filter_override_flag to be equal to 0 (instead of to be equal to ph_deblocking_filter_override_flag) when not present.
- Skip the signalling of ph_/slice_deblocking_filter_disabled_flag when pps_deblocking_filter_disabled_flag and ph/slice_deblocking_filter_override_flag are both equal to 1 and infer the value of ph/slice_deblocking_filter_disabled_flag to be equal to 0 under this condition.
For item 1, it was commented that it would be better to say “in the picture header or slice header” than “at the picture level or slice level”. Another participant said that the version from JVET-R0159 is more clear in terms of exactly how specific syntax element interact. This is an editorial matter that can be resolved by the editor.
It was commented and agreed that, as written, we cannot remove “in which slice_deblocking_filter_disabled_flag is not present” in the semantics of ph_deblocking_filter_disabled_flag.
Decision (sensibility cleanup): Adopt (with editorial matters to be resolved as noted above).
Quantization control signalling (6)
Discussion began here for JVET on 16 April at 0500 UTC (chaired by GJS, JRO, YKW).