Back to Search Document details
18th Meeting: Teleconference, April 2020 2020-04-08 03:37
AHG9: On signalling of deblocking control
Abstract
It is asserted that the deblocking control mechanism in the latest VVC text is pretty complicated, not straightforward, not easy to understand, and consequently prone to errors. This contribution proposes to change the deblocking signalling based on a 2-bit deblocking mode indicator in the PPS, summarized as follows:
JVET-R0072 AHG9: On signalling of deblocking control [Z. Deng, Y.-K. Wang, L. Zhang, K. Zhang, J. Xu (Bytedance)]

Discussed in AHG Session 1.8 on Tuesday 7 April at 2320 UTC (chaired by GJS).

It is asserted that the deblocking control mechanism in the latest VVC text is pretty complicated, not straightforward, not easy to understand, and consequently prone to errors. This contribution proposes to change the deblocking signalling based on a 2-bit deblocking mode indicator in the PPS, summarized as follows:

  • Replace the three PPS flags for deblocking signalling with a 2-bit deblocking mode indicator that specifies the following four modes: a) deblocking fully disabled and not used for all slices; b) deblocking used for all slices using 0-valued β and tC offsets; c) deblocking used for all slices using β and tC offsets explicitly signalled in the PPS; and d) deblocking further controlled at either picture or slice level.
  • The two flags in PH/SH are renamed to be ph/slice_deblocking_filter_used_flag and ph/slice_deblocking_parameters_override_flag, with the use flag specifying whether deblocking is used for the current picture/slice, and the override flag specifying whether the β and tC offsets are overridden by the values signalled in the PH/SH.

There was discussion of the various cases described in the document.

It was commented that deblocking is the only case where a disabling in the PPS can be overridden at the picture level. Some participants said this is OK, as it enables PPS sharing, and noted that we have such an override in HEVC. Others said this is the only place in the text that allows a disabling to be overridden.

It was agreed that there is no clear bug in the current design, although it does appear that there are a number of editorial bugs in the current semantics of the text.

No clear need for action was identified by the AHG on this, so the AHG did not recommend action.

Decisions
No clear need for action was identified by the AHG on this, so the AHG did not recommend action.
Citation