Incorrect equation in D.7.2 Subpicture level information SEI message semantics

[adybrowne]

Currently the MinCR constraint for subpictures for AU 0 is defined as:

– The sum of the NumBytesInNalUnit variables for AU 0 corresponding to the j-th subpicture sequence shall be less than or equal to

FormatCapabilityFactor * ( Max( AuSizeMaxInSamplesY[ 0 ] * OlsRefLevelFraction[ i ][ j ][ k ] ÷ 256, FrVal * MaxLumaSr * OlsRefLevelFraction[ i ][ j ] ÷ 256 ) + MaxLumaSr * ( AuCpbRemovalTime[ 0 ] − AuNominalRemovalTime[ 0 ] ) * OlsRefLevelFraction[ i ][ j ][ k ] ) ÷ ( 256 * MinCr ),

where MaxLumaSr and FormatCapabilityFactor are the values specified in Table 136 and Table 137, respectively, that apply to AU 0, at level sli_ref_level_idc[ i ][ k ], and MinCr, AuSizeMaxInSamplesY[ 0 ] and FrVal are derived as specified in A.4.2.

Two issues exist.

Firstly, there is a missing index [k] for OlsRefLevelFraction in the second part of the Max evaluation.

Secondly, both parts of the Max evaluation have a division by 256. This division is already covered by the division (256 * MinCr) at the end of the calculation.

Hence the corrected text for the calculation could be:

FormatCapabilityFactor * ( Max( AuSizeMaxInSamplesY[ 0 ] * OlsRefLevelFraction[ i ][ j ][ k ], FrVal * MaxLumaSr * OlsRefLevelFraction[ i ][ j ][ k ] ) + MaxLumaSr * ( AuCpbRemovalTime[ 0 ] − AuNominalRemovalTime[ 0 ] ) * OlsRefLevelFraction[ i ][ j ][ k ] ) ÷ ( 256 * MinCr )

Alternatively, the division by 256 could be moved from the last part of the calculation to the third use of OlsRefLevelFraction:

FormatCapabilityFactor * ( Max( AuSizeMaxInSamplesY[ 0 ] * OlsRefLevelFraction[ i ][ j ][ k ] ÷ 256, FrVal * MaxLumaSr * OlsRefLevelFraction[ i ][ j ][ k ] ÷ 256 ) + MaxLumaSr * ( AuCpbRemovalTime[ 0 ] − AuNominalRemovalTime[ 0 ] ) * OlsRefLevelFraction[ i ][ j ][ k ] ÷ 256 ) ÷ MinCr

A third alternative would be to extract the OlsRefLevelFraction[i][j][k] as a common factor as follows:

FormatCapabilityFactor * ( Max( AuSizeMaxInSamplesY[ 0 ], FrVal * MaxLumaSr ) + MaxLumaSr * ( AuCpbRemovalTime[ 0 ] − AuNominalRemovalTime[ 0 ] ) ) * OlsRefLevelFraction[ i ][ j ][ k ] ÷ (256 * MinCr)

[skupin]

Thanks Adrian, I can confirm the two issues and have a preference for your third proposed alternative fix.

[yk]

Good catch! Thanks Adrian for the fix and Robert for confirming! Will be fixed in the VVC attachment of JVET-W1004-v1 (taking the third alternative fix).