Picture coding: The use of a viewer model in source encoding

Abstract

A method is suggested for inserting viewer criteria directly into coding algorithms; any complex visual model may be used. The technique is applied to a DPCM-type coder, and a number of variations are compared on the basis of entropy, quality, and complexity. It is found that, using a simple one-dimensional filter model, the first-order entropy of the DPCM signal can be reduced by 30 percent for a high-detail picture with only a small reduction in picture quality. Furthermore, by means of a single threshold control, one can efficiently trade off bit-rate and picture quality over a large range for use in adaptive strategies.