JVET-Y0126 AHG10: VTM encoder configurations for tests targeting improved coding performance [D. Rusanovskyy, M. Karczewicz (Qualcomm), K. Andersson, R. Sjöberg, L. Litwic (Ericsson), P. Nikitin, G. Martin-Cocher (InterDigital), A. Wieckowski, J. Brandenburg, B. Bross (HHI)]

In this contribution, a set of encoder-only modifications and settings targeting improved compression efficiency of VTM reference software is proposed. Proposed method is implemented as minor encoder software change on top VTM-14.2 and encoder configuration files for RA and LD. It is reported that under the CTC, proposed method provides the following BD-rate gain vs. VTM-14.2 anchor:

SDR:

• RA: -2.66%/-6.24%/-6.13% with 168%EncT/96%DecT.
• LDB: -2.6%/-2.2%/-2.1% with 150%EncT/97%DecT.
• LDP: -3.0%/-2.6%/-2.8% with 171%EncT/101%DecT.

HDR RA:

• PQ: -2.6%/-4.1%/-2.1% (wPSNR) with 167%EncT/110%DecT.
• HLG: -3.3%/-3.9%/-5.3% (PSNR) with 165%EncT/90%DecT.

Against the VTM-15.0 anchor, the same RA results are achieved. For LDB and LDP configuration, BD-rate gains of -1.8%/-1.4%/-1.2% and -1.9%/-1.4%/-1.5% are reported, respectively.

It is proposed to modify VTM and:

• Add command-line option to control RDO cost factor for partitioning (default behaviour unchanged)
• Add command-line option to control fast decision for TT from BT (default behaviour unchanged)
• Speed-up RPR handling (bit-exact change)

It is further proposed to:

• include configuration files for tests targeting improved coding performance, for example for benchmarking purposes. Such files would be put into a separate configuration folder.

No changes to VTM CTC are requested.

It was noted that some changes may be applied to ECM as well.

Decision (SW): Include the proposed VTM changes into next VTM release.

Decision (SW): Include proposed configuration files in a separate folder in the next VTM release (no change to CTC).

Further study, including applicability to ECM, was encouraged.