Multifractals and texture classification

In image analysis, texture recognition and discrimination is a major task. Fractal geometry is often used in this particular application for its capability in characterizing texture using a single parameter, that is, the fractal dimension. However, fractal analysis presents the major drawback to not be able to fully characterize a given surface; as a matter of fact, quite different textures have, often, the same fractal dimension. In the paper, a methodology for texture analysis, characterization, and classification is presented that is aimed at overcoming such drawback. To this end, the authors developed a technique based on the estimation of the multifractals features of a surface, i.e., the function D(q) of each texture patch, and by using such information they performed an accurate classification of different textures. Results obtained on standard Brodatz texture are reported and discussed