Spec and VTM mismatch on chroma QT split at picture boundary

[hanhuang]

In VTM-8.0, when x0 + cbWidth is greater than pic_width_in_luma_samples and y0 + cbHeight is greater than pic_height_in_luma_samples and cbWidth is equal to MinQtSizeC in the chroma tree, quad split is derived as implicit split. But in the VVC draft 8, both horizontal binary split and quad split are allowed.

If align spec with VTM, the text change is as follows, with added text highlighted:

– if all of the following conditions are true, allowBtSplit is set equal to FALSE

– x0 + cbWidth is greater than pic_width_in_luma_samples
– y0 + cbHeight is greater than pic_height_in_luma_samples
– cbWidth is greater than minQtSize and treeType is not equal to DUAL_TREE_CHROMA

– if all of the following conditions are true, allowBtSplit is set equal to FALSE

– x0 + cbWidth is greater than pic_width_in_luma_samples
– y0 + cbHeight is greater than pic_height_in_luma_samples
– cbWidth is greater than (minQtSize* SubHeightC / SubWidthC) and treeType is equal to DUAL_TREE_CHROMA

[bbross]

Thanks for reporting!

Can someone confirm the suggested fix?

[hanhuang]

This is solved by adoption of JVET-R0131.
Replying to bbross:

Thanks for reporting!

Can someone confirm the suggested fix?

[bbross]

Thanks so this is fixed in D9.