Texture of a fatigue crack surface is strictly related to crack growth rate. Cracks in specimens from aluminum alloy were studied. Two types of information were used: SEM images of fracture surfaces, and 3D reconstructions of fracture surface morphologies. Sets of equidistantly focused images obtained by an optical microscope served as the basis for 3D reconstruction. Multiparametric fractal analysis was applied to characterize crack surfaces. A vector of fractal features represented each image or 3D reconstruction of selected locations of fracture surfaces. For estimating fractal characteristics, the box-counting method in 3D was used in all cases, . Multilinear regression was used to express the relation between crack growth rate and feature vectors, with satisfactory results for both crack surface representations.