Trellis coded residual vector quantization with application to image coding

Abstract

This paper describes a vector quantization approach that combines the residual Vector quantization (RVQ) with trellis-coded vector quantization (TCVQ). The resulting quantizer is referred to as trellis-coded residual vector quantizer (TCRVQ). An entropy-constrained (EC) version is also developed and tested on synthetic as well as real sources. Performance of the entropy-constrained version of our quantizer provides improvement over the ECRVQ (entropy-constrained residual vector quantization) and ECTCVQ (entropy-constrained trellis-coded vector quantization) for a fixed level of complexity. Simulation results for image coding indicate that the new scheme achieves 0.8 dB improvement over the comparable RVQ approaches in terms of signal-to-noise ratio (PSNR).