JVET-AP0057 AHG9: Some potential improvements for the FGRC SEI message [Y.-K. Wang, J. Xu (Bytedance)]

This contribution proposes some potential improvements for the film grain regions characteristics (FGRC) SEI message in JVET-AO2032-v2. Besides proposed changes, this contribution also raise some open discussions on some aspects of the FGRC SEI message.

Below is a summary of the proposed changes and the open discussions:

1. On signalling of fgr_region_interval_max[ i ][ c ]
2. On a constraint requiring the presence of the four VUI-type FGC parameters
3. On interface text: inference of FGC picture width and height for VVC
4. On one presence flag for FGC picture width and height and bit depth syntax elements
5. On film grain parameters for chroma components
6. Disallow any two explicitly indicated regions to be the same
7. Disallow not using an entire set of modelling parameters for a film grain model
8. Disallow not using the modelling parameters for any intensity interval
9. Open discussion 1: Apply to decoded picture or cropped decoded picture?
10. Open discussion 2: Use InPic or (cropped) decoded picture in the semantics?

It was questioned how important it is to align FGC with FGRC. It is noted that they have already drifted, and it was suggested that we can deviated.

Agreed to items 1, 2, 5. 7

Further discussion requested on item 3, 4.

For item 5, it was questioned if ChromaFormatIdc can be used instead of the proposed three components flag.

No action on item 6. 8

For item 9, it was agreed to apply to cropped decoded picture. Text is delegated to editors.

For item 10, it was agreed to use InPic in the semantics. Text is delegated to editors.

Further discussed at 1710 on Wednesday 29 April chaired by J. Boyce.

Further study on item 3.

Agreed on item 4 with the if condition moved to be the first set of syntax elements in the syntax table after the cancel flag.

JVET-AP0057 AHG9: Some potential improvements for the FGRC SEI message