Incorrect subpicture boundaries

[bheng]

In equation 113, I believe SubpicRightBoundaryPos and SubpicBotBoundaryPos should be derived using the actual picture size from the PPS, not the maximum picture size from the SPS:

SubpicRightBoundaryPos = Min( sps_pic_width_max_in_luma_samples - 1, ... )

SubpicBotBoundaryPos = Min( sps_pic_height_max_in_luma_samples - 1, ... )

These boundaries are used in several TMVP equations, such as the following:

rightBoundaryPos = sps_subpic_treated_as_pic_flag[ CurrSubpicIdx ] ? SubpicRightBoundaryPos : pps_pic_width_in_luma_samples - 1 (585)

When not present, the value of sps_subpic_treated_as_pic_flag[ i ] is inferred to be equal to 1. So, the subpicture boundaries need to be derived correctly above, even if subpictures are not used.

[yk]

Thanks! I agree. And since the equation is located in the SH semantics (it was originally located in the SPS semantics), in any case it is safe to use the actual picture size from the PPS.

[bbross]

Fixed as discussed/suggested in JVET-T0110-v2.