MIsmatch with vtm on cu chroma qp offset enabling

[forayr]

The specification decides to enable cu_chroma_qp_offset tools only at the PPS level, whereas VTM checks also a flag at the slice level (after 'slice_joint_cbcr_qp_offset').
Is this a VTM issue ?

[delagrangep]

Thanks for raising the issue.

This code is inherited from HEVC, and this is rather a spec issue: we make use of "cu_chroma_qp_offset_enabled_flag" which is undefined in VVC Draft.

Actually in JVET-O1168 we forgot to port the slice-level flag from HEVC, just before "if( deblocking_filter_override_enabled_flag )":

if( chroma_qp_offset_list_enabled_flag )

cu_chroma_qp_offset_enabled_flag u(1)

I apologize for having missed it.

We may discuss this issue at the next meeting, however the meeting notes seem to suggest alignment with HEVC in the discussion of JVET-O0737: "Hypothetically, we could just basically do what is in HEVC (“variant 0”)", that "variant 0" being what was adopted as JVET-O1168.

I can provide the spec fix if everyone is ok.

[delagrangep]

Suggested changes to JVET-O2001-vE

[bbross]

Thanks for reporting and suggesting a fix.

This will be fixed in JVET-P2001-vA.