Automatic segmentation of different types of tissue from medical images of several sensing modalities is of great importance for clinical and research applications. In this paper, we propose a segmentation methodology based on a multifractal approach. We present different alternatives for estimating local fractal exponents in these images, as well as their global distributions –the multifractal sperctrum. We generate new images by means of grayscale mapping of these local an global computed values. The obtained results are quite promising as a tissue differentiation tool, and therefore are suitable to carry out automatic segmentation of abnormalities in medical images.Sociedad Argentina de Informática e Investigación Operativa