A new technique for adaptive scalar quantization of image sub-band coefficients

A new method for adaptive scalar quantization of image sub-band coefficients has been developed. This method is based on a phase dispersion technique originally developed for the design of pulse compression radar signals. For this method, the coefficient magnitudes and signs are coded separately. The magnitudes are scanned into a set of 1-D same orientation, cross-band vectors. Each vector is processed using a non-linear transformation. The result of this operation is a vector whose components are approximately identically distributed. Therefore, only one quantizer and one entropy coder are needed to code each vector. The technique offers the performance of a fully adaptive compression scheme, but with modest overhead and computational requirements. This paper provides an overview of the method as well as experimental results obtained through simulation.