[Editorial] grain_scaling_minus_8 semantics

[peterderivaz]

grain_scaling_minus_8 is used during film grain synthesis for both luma and chroma, so correct description that used to say it only applied to chroma.

[imoccaga]

I guess the code use grain_scaling_minus_8 during film grain synthesis for both luma and chroma, so can we classify this as a spec fix to match the code and the original intent ?

If so, this seems a safe fix because if an existing decoder uses grain_scaling_minus_8 only for chroma it would generate an incorrect film grain synthesis value and it would be a non compliant / broken decoder.

What do you think ?

[peterderivaz]

@imoccaga Yes, this is a safe fix as it only affects the semantics. It fixes the description of the syntax element but does not change normative behaviour.

The normative decoding process for film grain is in a separate part of the specification and matches the code.

[imoccaga]

The fact that a change affects only the semantic does not make it safe by construction.

Also, do we have to make the same change in AFGS1 spec too ?

[peterderivaz]

The fact that a change affects only the semantic does not make it safe by construction.

Agreed. I was trying to point out that in this case the specification is currently internally inconsistent as there is a separate part that goes into more detail about how to apply the grain, and the more detailed part does this correctly.

Also, do we have to make the same change in AFGS1 spec too ?

Yes, this text is also present in https://github.com/AOMediaCodec/afgs1-spec/blob/main/07.bitstream.semantics.md so should be updated there as well.