Digital Color Image Compression in a Perceptual Space.

摘要

Simple mathematical models are developed from the physiological and psychophysical traits of the human visual system. Expressions for the statistical characterization of these models are obtained. When used as a preprocessor, the models are shown to produce images which are statistically compatible with the underlying assumptions necessary to solve the parametric rate-distortion equations. The derived power spectrum equations were used to code black and white and color images with a quality superior to previous results. In addition, it is shown that the preprocessor produces a 'perceptual space' in which normalized mean square error is an effective image quality measure. (Author)