In this paper, we present interesting correlations between subjective scores of image complexity and a series of computational measurements of geometrical properties of presented images. A set of 49 images was applied as a systematic methodology to characterize the human perception of its complexity. The experiment requested subjects to assign scores (0-10) to each images according to their subjective judgment of image complexity. A total of thirty five subjects with ages ranging from 30-50 years, all of them post-graduated in exact sciences, were considered. A series of measurements of geometrical properties (mean, standard deviation, entropy and lacunarity) of the presented images were obtained using programs run in the SCILAB 2.6 environment. All the 49 images were converted into matrices and averages and standard deviations of the luminosity in such matrices, as well as their respective entropies and lacunarity, were estimated using standard image processing methodology. The obtained results point to remarkable correlations, paving the way to further investigations upon visual complexity using modeling based on data mining.

image complexity, image perception, multivariate imaging techniques.