JVET-R0243 AHG9: 4:4:4 vs. 4:2:0 bit-rate in VTM [S. Keating, A. Browne, K. Sharman (Sony)]
This was discussed in AHG Session 1.5 (chaired by GJS, YKW, JRO & JB).
This contribution compares bit rates for 4:4:4 and 4:2:0 encoding. It is for information/discussion only.
Another contribution, JVET-R0244, proposes changes to the CpbVclFactor and MinCrScaleFactor for Main 4:4:4 10 profile. JVET-R0244 proposes to specify that the maximum bit-rate of 4:4:4 should be twice the maximum bit-rate of 4:2:0 (instead of 2.5 as currently specified, and as in AVC and HEVC).
Comments:
In some cases, unless low-pass filtering is applied, the bit rate is higher. However, on the other hand, low-pass filtering to chroma seems not good as it blurs the chroma.
Chroma QP offset or lambda adjustment are another ways of adjusting the bit-rate balance.
A reason for having some extra bit rate header room for 4:4:4 is that the quality expectation for 4:4:4 is higher, and the GOP length or intra refresh period may be shorter.
Sometimes there is RGB coding for 4:4:4, which is generally less efficient than YCbCr.
Discussion stopped here for AHG Session 1.5 Tuesday 7 April at 1500 UTC.