Mismatch with spec on modeTypeCondition derivation.

[hmjang]

There is unnecessary condition which seems like mismtach for modeTypeCondition derivation.

In specification, the below condition is not necessary. this condition seems not at VTM-6.0rc1.

• cbWidth * cbHeight is equal to 32 and MttSplitMode[ x0 ][ y0 ][ mttDepth ] is equal to SPLIT_BT_HOR or SPLIT_BT_VER

This condition rather seems to allow 4x2 chroma intra block. Becasue for example when the current block is 8x4 luma block, modeType is derived as MODE_TYPE_INTRA, even though the corresponding chorma block is 4x2.

Lastly, this case;8x4 and 4x8 block which modeType is MODE_TYPE_ALL cannot happended by the restricted modeType of parent CU.

[bbross]

Can somebody confirm that:
a) this introduces 4x2 and 2x4 intra chroma blocks
b) this is a mismatch with VTM
c) removing this condition would fix the spec according to what we intended to do when adopting JVET-O0050 and align spec and VTM

[zhaoyin]

a) this introduces 4x2 and 2x4 intra chroma blocks

--no

b) this is a mismatch with VTM

--yes, VTM does not have this condition check.

c) removing this condition would fix the spec according to what we intended to do when adopting JVET-O0050 and align spec and VTM

--JVET-O0050 was proposed to solve small intra chroma blocks in YUV4:2:0 format, and this condition is just reduandant in this case.

[bbross]

Thanks for reporting and confirming.

The condition will be removed in JVET-P2001-vD.

[zhaoyin]

The condition is useful for YUV4:2:2 format, so don't remove it. This issue has been solved by adopting JVET-P0063/P0406/P0520/P0537/P0575 (fix local dual tree on color format)

Replying to bbross:

Thanks for reporting and confirming.

The condition will be removed in JVET-P2001-vD.

[bbross]

Thanks for pointing that out, re-integrated the condition.