JVET-R0219 Alternative block size conditions for BDPCM [K. Unno, K. Kawamura, S. Naito (KDDI)]
In the last meeting, it was adopted that chroma BDPCM for 4:2:0 and 4:2:2 format. In the current spec, BDPCM applied CU can be separated into multiple TUs depending on the max transform skip size and the max TU size, in contrast with BDPCM applied CU always has only one TU in the VVC D7. In this contribution, it is proposed to change the block size condition of BDPCM. Two alternatives are proposed. Method 1 proposes to add a restriction that chroma BDPCM can be applied only if the CU is not separated into multiple TUs. Method 2 proposes to relax block size conditions that BDPCM can be applied regardless of separating CU to multiple TUs for not only chroma but also luma.
Experimental of Method 1 with CTC QPs show BD-rates for luma are ‑0.01%/0.02%/0.04% (AI/RA/LD) for Class F, ‑0.01%/‑0.02%/‑0.09% (AI/RA/LD) for Class TGM.
Experimental of Method 1 with Low QPs show BD-rates for luma are 0.00%/0.01%/‑0.01% (AI/RA/LD) for Class F, ‑0.02%/‑0.02%/‑0.01% (AI/RA/LD) for Class TGM.
Experimental of Method 2 with CTC QPs show BD-rates for luma are ‑0.09%/‑0.06%/0.13% (AI/RA/LD) for Class F, 0.01%/‑0.02%/0.03% (AI/RA/LD) for Class TGM.
Experimental of Method 2 with Low QPs show BD-rates for luma are‑0.02%/0.04%/‑0.04% (AI/RA/LD) for Class F, ‑0.02%/0.02%/‑0.09% (AI/RA/LD) for Class TGM.
During the discussion, it is pointed out that there is not a design problem, and it is also not asserted as inconsistent that for certain block sizes BDPCM is restricted for luma, but not for chroma.
No action was necessary on this.