JVET-AF0010 JVET AHG report: Encoding algorithm optimization (AHG10) [P. de Lagrange, A. Duenas, R. Sjöberg, A. Tourapis (AHG chairs)]

Related contributions

A total of 5 contributions, not including cross-checks, are identified relating to AHG10, and summarized in the following sections.

DMVR encoder control

JVET-AF0057 – EE2-3.5: DMVR with robust MV derivation

Though this contribution in an EE2 test, it contains an encoder-only aspect (test 3.5b), maxing out the RD cost of coding modes where DMVR is likely to create annoying artefacts. The BD-rate impact can be up to around 2% for high QPs (the normative solution seems to have both much less BD-rate impact and better subjective quality).

Adaptive resolution

JVET-AF0058 – AHG12: GOP-based RPR encoder control for ECM

This contribution proposes to port the GOP-base RPR encoder control feature available in the VTM to ECM. The gains seem to be reduced by more than half, but are still significant (around 1.5%; more than 3% in class A). It is said that subjective quality is also improved.

Local QP optimization

JVET-AF0089 - AhG10: Lagrange multiplier and QP adaptation at CTU-level for VVC

This contribution proposes a method that tries to give more quality weight to local areas that are likely to be reused in other picture (through temporal prediction). It does so by estimating a “distortion propagation factor” with a fast 1^st inter pass. BD-Rate of around -4% is reported for LD and -0.25% for RA.

This is probably related to the “block importance mapping” feature present since VTM-16.0 (but not enable that default), that makes use of MCTF-derived metrics, and gives around -2% BD-rate for both RA and LD.

Fast methods

JVET-AF0111 – AHG10: MTT split modes early termination

This contribution further studies an early termination method that reduces encoding runtime for ECM and VTM (JVET-AE0057, adopted in ECM CTC, and VTM but not in CTC), by skipping MTT splits when the no-split cost of 64x64 CU is above a threshold. Other bit depths and larger QP range are tested, and some threshold adjustments are made. Encoding times of -3% to -4% are reported, with a BD-rate difference in the range of +0.02% to +0.06%, the tradeoff being improved compared to JVET-AE0057 for high QP values.

Lamba optimization

JVET-AF0122 - AhG10: Lagrange multiplier optimization for chroma ALF and CCALF

As in JVET-AD0136 and JVET-AE0123, this contribution proposes to adjust the lambda for the RDO of chroma ALF when luma ALF coefficients are to be transmitted, to account for a reduced overhead of the combined signalling, and also encourage combined signalling (resulting in more chroma filter diversity). Chroma gains are reported.

Recommendation

The AHG recommended that the related input contributions be reviewed and to further continue the study of encoding algorithm optimizations in JVET.