Concentric-shell partition vector quantization with application to image coding

A method is described for designing a vector quantizer by first partitioning the image vector space into concentric shells and then searching for the smallest possible codebook to represent it, while adhering to the visual perceptive qualities such as edges and textures in the image representation. The coding method involves decomposition of the image vector space into relatively independent components, each with a small dynamic range. Optimal partitioning amounts to optimally selecting the set of shell radii, using a modified form of simulated annealing. The results show that the concentric-shell partition vector quantizer outperforms the gain-shape vector quantizer in terms of computational complexity and coded image quality