Two greedy tree growing algorithms for designing variable rate vector quantizers

This paper presents two variable rate vector quantization schemes for image compression. The idea is to designate more bits to regions of an image that are active or difficult to code, and fewer bits to less active regions. The proposed schemes design variable rate tree-structured vector quantizers by growing the tree one node at a time. Basing the decision to split on some min-max-like criteria other than the conventional mean-squared-error (MSE) measure, these two schemes aim to provide reconstructed images with less noticeable edge degradation. A comparison with a previously proposed scheme of Riskin and (see IEEE Trans. Signal Process., vol.39, no.11, p.2500-2507, 1991) shows that the proposed schemes have better performance in terms of image visual quality with reduced implementational complexity