Texture representation through fractal singularity spectrum

In this paper we deal with multifractal theory and its application to image analysis. Based on multifractal theory fundamentals, we illustrated an estimation algorithm for singularity spectrum that fits the case of digital images, that is a discrete set of points resulting from sampling a continuous function. The singularity spectrum is defined, and its interesting properties are presented and discussed, together with a novel estimation technique. Such a method is based on the concept of box-dimension and box-counting, but innovations have been introduced. Finally, some experimental results are presented and discussed.