Progressive Vector Quantizations with Applications to Medical Images.

Progressive vector quantization is an image compression scheme invented to precisely control the coding error of every pixel. Unlike most image coding algorithms that minimize a global distortion measure but have no control on the individual pixel errors, progressive vector quantization can cooperate with a fidelity criterion that applies to single pixels as well as local regions. The user can even choose one or a few error- patterns that are forbidden in the coded images. When the zero-error is our selection, it can also produce error-free coded pictures. This coding scheme is particularly useful for the applications where high fidelity is required such as medical image compression. Our simulations adopt a simple absolute error criterion which only allows a 0.8% or less coding error at every pixel; still a compression ratio around 8:1 has been achieved.