Redundancy processing on qPY_PREV derivation

[abe.kiyo]

In 8.7.1 Derivation process for quantization parameters, there is the following processing.

1. The variable qPY_PREV is derived as follows:
   • If one or more of the following conditions are true, qPY_PREV is set equal to SliceQpY:
     • The current quantization group is the first quantization group in a slice.
     • The current quantization group is the first quantization group in a tile.
     • The current quantization group is the first quantization group in a CTB row of a tile and sps_entropy_coding_sync_enabled_flag is equal to 1.
   • Otherwise, qPY_PREV is set equal to the luma quantization parameter QpY of the last luma coding unit in the previous quantization group in decoding order.

....

4. The predicted luma quantization parameter qPY_PRED is derived as follows:
   • If all the following conditions are true, then qPY_PRED is set equal to the luma quantization parameter QpY of the coding unit containing the luma coding block covering ( xQg, yQg − 1 ):
     • availableB is equal to TRUE.
     • The current quantization group is the first quantization group in a CTB row within a tile.
   • Otherwise, qPY_PRED is derived as follows: qPY_PRED = ( qPY_A + qPY_B + 1 ) >> 1 (1140)

qPY_PREV is set equal to SliceQpY by the condition of red text for WPP. However, according to the blue text, qPY_PRED is set equal to QpY of above CU without using qPY_PREV for both WPP and non-WPP. To avoid the confusion in WPP specification, I think the processing of red text should be removed.

[bbross]

Good catch, thanks!

The redundancy was carried-over from HEVC when WPP was added. In HEVC, the carry-over from the last CTU of the CTU line above was only disabled for WPP.
In VVC, this was introduced independently of WPP by the adoption of JVET-M0685.

Hence, I agree that the red text can be removed and this will be fixed in the first version of D10.