| JCTVC-M0205 MV-HEVC/SHVC HLS: On inter-layer sample and syntax prediction indications [M. M. Hannuksela, K. Ugur (Nokia)]
Discussed in BoG; see BoG report M0450.
Resolved by M0457-v2.
JCTVC-M0457 MV-HEVC/SHVC HLS: On inter-layer sample and syntax prediction indications (combining aspects of JCTVC-M0205 and JCTVC-M0046) [M. M. Hannuksela (Nokia), J. Boyce (Vidyo)]
Discussed in BoG; see BoG report M0450.
After offline work aiming at a harmonized design among the proponents, M0457-v2 was provided and reviewed:
-
Sequence-level signalling is in VPS (not in SPS)
-
A separate loop than the layer dependency signalling, 2 types of ILP (sample and motion) for both SHVC approaches.
-
If in the future only one of the two SHVC approaches is supported, then a change can be done to support only that approach.
-
Supports a hypothetical scenario of having both SHVC approaches supported even within a coded video sequence, but cannot be combined within a layer in a CVS.
-
Four variables that are currently used in the SHVC test model text are derived based on the syntax elements.
-
Slice-level signalling
-
First part is the same as in proposal#3 of M0205.
-
Second part: Two flags for ILP of sample enabled and ILP of sample only. And another flag for ILP of motion.
-
Derivation of TMVP collocated picture (same as in proposal#3 of M0205)
-
Derivation of RPS and RPL, to include only the lower-layer pictures that are used for ILP of sample. It was agreed that, if to do this, only do it for RPL, not for RPS.
-
Remove the _mfm_ flag in the SHVC test model (same as in proposal#4 of M0205).
-
For the TextureRL based approach, use the slice level flag to derive a flag for syntax controlling in the CU syntax.
Sequence-level signalling in the VPS seems more flexible than in the SPS. Also, it would avoid increasing the number of SPSs.
Decision (affirming BoG M0450 recommendation): Agreed.
Dostları ilə paylaş: |
