Entropy-based representation of image information

Loss of information in images undergoing fine-to-coarse transformations is analysed by using an approach based on the theory of irreversible processes. In the case of grey level images, entropy variation along scales is used to characterize basic, low-level information and to identify perceptual components of the image, such as shape and texture. Here an extension of the approach to colour images is proposed. Spatio-chromatic information is defined, which depends on cross-interactions between the different colour channels. Examples illustrating the use of spatio-chromatic information are presented, related to pattern recognition and active vision. (C) 2002 Elsevier Science B.V. All rights reserved.