Space variant resolution ADCT still image coding

Abstract

Starting from the principle that the distorsion measure, representing the cost payed by the users for lossy reproducing an image, takes its maxima in correspondence of local minima of Shannon's Entropy, a space variant resolution coding method matched to the local content of the image is presented. A fidelity criterion based on the estimated kurtosis function of the image is then applied to drive the quantization phase of an ADCT coder. Finally, an enhanced version of the ISO/CCITT hierarchical progressive transmission mode is proposed and its perfomance evaluated on the basis of reference images.