Hybrid Video Coding with Trellis-Coded Quantization

In state-of-the-art video coding, the prediction error signals are transmitted using transform coding, which consists of an orthogonal transform, scalar quantization, and entropy coding of the quantization indexes. We show that the coding efficiency of transform coding can be improved by replacing scalar quantization with trellis-coded quantization (TCQ) and using advanced entropy coding techniques for coding the quantization indexes. The proposed approach was implemented into the first test model (VTM-1) of the new standardization project Versatile Video Coding (VVC). Our coding experiments yielded average bit-rate savings of 4.9% for intra-only coding and 3.3% for typical random access configurations, where bit-rate savings of 3.5% (intra-only) and 2.4% (random access) can be attributed to the usage of TCQ. These coding gains are obtained at a 5-10% increase in encoder run time and without any change in decoder run time.