Image coding using pyramid vector quantization of subband coefficients

This paper presents an improved algorithm using pyramid vector quantization with subband decomposed images. Specifically, the use of large vector dimensions and different dimensions for each subband yields significant improvement over previously reported results. Simulations reveal compression performance comparable to JPEG using a purely fixed-rate code, which has less hardware complexity and greater error resiliency. A comparison between product and inner pyramid VQ using statistical analysis of subband data and simulations demonstrates that product pyramid VQ is better suited for subband coding