JVET-Q0137 CE3-related: Rice Parameter Derivation with Unified Lookup Table [T. Tsukuba, M. Ikeda, Y. Yagasaki, T. Suzuki (Sony)]

This contribution proposes a rice parameter derivation using a unified lookup table for regular residual coding and transform skip residual coding to simplify CE3-1.1-subset3.

It is reported that, at common QP condition, proposed method provides average BD-rate differences of (-0.01%, 0.00%, -0.02%) for AI, (-0.02%, -0.01%, 0.02%) for RA, and (-0.02%, -0.01%, 0.02%) for LB; For ClassF, (-0.16%, -0.17%, -0.04%) for AI, (-0.18%, -0.03%, -0.17%) for RA, and (-0.13%, -0.58%, 0.62%) for LB; For ClassTGM, (-0.39%, -0.30%, -0.32%) for AI, (-0.19%, -0.20%, -0.16%) for RA, and (-0.20%, -0.35%, -0.26%) for LB.

It is reported that, at low QP condition, proposed method provides average BD-rate differences of (-0.07%, -0.10%, -0.12%) for AI, (-0.02xx%, -0.02yy%, -0.04zz%) for RA, and (-0.02%, -0.02%, -0.01%) for LB; For ClassF, (-1.41%, -0.90%, -0.90%) for AI, (-1.84%, -0.73%, -0.86%) for RA, and (-0.63%, -0.55%, -0.70%) for LB; For ClassTGM, (-1.58%, -1.44%, -1.45%) for AI, (-2.13%, -1.40%, -1.39%) for RA, and (-1.44%, -1.17%, -1.20%) for LB.

It is reported that, at lossless condition, average bitrate differences of -5.08% for AI, -3.73% for RA, and -2.58% for LB; For ClassF, -3.97% for AI, -3.06% for RA, and -2.78% for LB; For ClassTGM, -4.35% for AI, -6.33% for RA, -6.57% for LB.

In v3, simulation results for ClassA2/RA at low QP condition were updated.

There was no need to consider this after the CE3 decision.