Accurate textural characterisation using fractal techniques

It has been demonstrated that the fractal dimension is not sufficient to completely categorise texture. However, fractal techniques do have advantages over statistical and structural methods, by providing a framework more suited to natural textures. Many natural textures show fractal-like properties of statistical, if not deterministic, self-similarity. Natural textures exhibit only a limited range of scale for self-similarity. If a fractal dimension is stable over a wide range of scale then it is reasonable to term the textural surface as fractal. A number of approaches are described, using 50 natural textures and sets of artificially generated fractal surfaces, in order to determine robust texture characterisation.