On Scalar Quantizer Design with Decoder SideInformation

摘要

In this work, we investigate the design ofrnscalar quantizers for the Wyner-Ziv problem, where therndecoder has access to a noisy copy of the source. Necessaryrnconditions are given for the design of optimum quantizersrnwith a fixed number of partitions for a criterion consistingrnof a linear combination of distortion and rate. A simplerniterative algorithm is developed to tackle the new quantizerrndesign problem. The new algorithm generalizes the well-knownrnLloyd type I algorithm. The new approach will,rnin general, yield different scalar quantizers than therntraditional approaches and these new quantizers providernsome advantages. We also develop approximations tornsimplify quantizer design for high rate cases. With suchrnapproximations, the traditional optimal fixed-rate andrnentropy-constrained scalar quantizers are shown to remainrnoptimum under certain conditions.