Multifractal texture classification of images

This paper presents a method for measuring the generalized information content in grey level images. This measure involves the use of a multifractal distribution function. The multifractal measure is based on the generalized entropy and correlation functions to determine the entropy distribution. The multifractal distribution function partitions the image into subsets, each of which has a different entropy. While the idea of obtaining a generalized entropy of a natural image has always been sought for in the literature, this information content has not up to now been described in terms of a multifractal distribution function, as it is in this paper. We report the Hausdorff fractal dimensions for Lena and the baboon of 2.5958, and 2.6562, respectively. The multifractal entropy distribution function f(/spl alpha/) shows a slightly wider breadth for the baboon as compared to Lena, indicating the baboon contains a higher degree of non-uniformity.