Mathematical morphology allowed quantifying the spacial arrangements of soil material and pores from thin section where size, shape and connective of pores are used to classify soil structure. However, the selection and interpretation of the morphological parameters that best characterize soil structure is still a subject of research. In this context, the application of multifractal analysis is used to characterize soil porosity and structure. This technique considers de inner variations of a complex system including density variations in the measures and the result is a graph well known as multifractal spectra. The objective of this research was to characterize the soil pore system under natural forest in the Caetetus Ecological Station (São Paulo, Brazil) using multifractal analysis and correlate multifractal parameters with soil attributes (particle size distribution, organic matter content, macroporosity and entropy of soil pore size distribution). The multifractal analyses of pore system were made with soil images by the Multifractal Analysis and Scaling System Algorithm – MASS (version 3.0). The soils horizons showed different shapes and symmetries by multifractal analysis indicating differences of pores distribution. The multifractal parameters of symmetry f(alpha)ratio, (alpha)ratio and delta(alpha) showed significant differences (p < 0,05) for A, E and B horizons. However, values of D0, D1 and D2 were significant different only for the B horizons (p < 0,05). Thus, the multifractal analysis was able to distinguish soils with different texture and structure. The correlation between multifractal parameters and soil attributes permitted the prediction of the multifractal parameters by Generalized Linear Model.