JVET-R0108 AHG9: Decoding Capability Information and PTL Signalling [S. Deshpande, J. Samuelsson, A. Segall, P. Cowan (Sharp)]
Discussed in AHG Session 1.5 Tuesday 7 April at 1410 UTC (chaired by GJS, YKW, JRO & JB).
Following is proposed related to DCI and PTL signalling:
- Proposal 1: It is proposed that dci_max_sublayers_minus1 syntax element be removed and instead those bits and the reserved zero bit be used for the syntax element dci_num_ptls_minus1.
- Proposal 2: It is proposed to rearrange the syntax elements in profile-tier-level signalling structure such that general_level_idc syntax element, which is unconditionally signalled, is first in the structure and the other conditional signalling, which is all based on profileTierPresentFlag is together, thus requiring only a single if check.
- Proposal 3: It is proposed to conditionally signal sps_ptl_dpb_hrd_params_present_flag only when sps_video_parameter_set_id is not equal to 0.
Proposal 1:
It was commented that the added semantics constraint on vps_num_ptls_minus1 has a problem.
It was suggested to use 4 reserved bits instead of having 8 bits for the number of PTL structures.
Comment: The current extension mechanism for DCI is a bit heavy.
Suggestion: Reserve a value of dci_num_ptls_minus1, e.g., 15.
AHG recommendation (cleanup): Remove the dci_max_sublayers_minus1 SE, but to use 4 reserved bits (at the begin of the DCI syntax) instead of having 8 bits for the number of PTL structures (as proposed), and reserve the value 15 of dci_num_ptls_minus1.
Proposal 2:
Comments: This makes level goes before profile and tier, while the interpretation of level typically depends on profile.
Comments: The GCI syntax structure can be of variable length. The proposed change makes level at fixed position.
No action was taken on this item.
Proposal 3: See notes in 6.1.6.1.
After discussion of this, at the end of AHG session 1.5, the information contribution JVET-R0243 was discussed. See the notes for that contribution, which have been moved to section 4.5.