Redundancy-complexity tradeoff of multiresolution coding for successively refinable sources

In this paper, we investigate the rate-distortion performance of multiresolution encoding algorithms for successively refinable sources, including block codes and trellis codes. Our random coding arguments not only support the results of simulation-based experiments, such as tree structured vector quantization (TSVQ) and multistage trellis coded quantization (MSTCQ), but also provides some information theoretic considerations to understand the efficiency of such existing algorithms. In fact, the redundancy analysis that we developed is a practical generalization of the Shannon source coding theorem to the more concrete cases.