A Biologically Motived Fuzz-like Measure for Complex Structures Perception

Abstract

In contrast to the perceptual capability of artificial systems, the biological perception of spatial patterns is a continuous cognitive process. In particular, the visual system of primates has a space-variant nature where the resolution is high on the fovea and decreases continuously to the periphery of the visual field. Moreover, the pattern perception and recognition may change, also continuously, when orientation and depth changes. An interesting aspect is that the perceptual performance needs to increase when the structure in recognition gets more complex in terms of irregular spatial contents (asymmetries). Based on these properties, we introduce a computational measurement procedure where the asymmetries are "continuously" quantified using intersections among partially fuzzy images. The asymmetries are quantified using the first gradient moment from the Gradient Pattern Analysis methodology. In this application, the first gradient moment is a fuzzy parameter whose fuzzy deviation is set in the same level of biological perceptual uncertainty. The performance of our approach is tested over texture variation perception in SAR (Synthetic Aperture Radar) images and the results show that this measure can be useful for real-time machine navigation and, in a more general sense, for biologically motivated morphology research.