Intra reference sample subsitution process uses incorrect indexing (8.4.5.2.9)

[swarrington]

Section 8.4.5.2.9 reference substition process has a bulleted list describing the process for handling when "at least one but not all samples refUnfilt[ x ][ y ] are marked as "not available for intra prediction"".

There appears to be an error in bullet 2:

2. For x = −1 − refIdx, y = −( refH − 2 )..1 + refIdx, when refUnfilt[ −x ][ −y ] is marked as "not available for intra prediction", the value of refUnfilt[ −x ][ −y ] is set equal to the value of refUnfilt[ −x ][ −y + 1 ].

x is defined as the left boundary pixel position (x = -1-refIdx). However, both the "not available for intra prediction" and substitution processes use "-x" indexing. ie: after x is converted to -x, it is pointing within the TU rather than the intended left boundary.

I believe the intention is for all the indexing in this bullet to use "x" rather than "-x".

[fbossen]

Look like this issue was introduced in JVET-S2001vE when we changed the bounds for y from y=a..b to y=-a..-b to address issue #1218. While y was correctly changed to -y when referencing refUnit, x should not have been changed to -x.

[yk]

Thanks! This will be fixed in the VVC attachment of JVET-Y1004, the ITU-T VVC v2 text, and the ISO/IEC VVC v2 FDIS text.