Properties of independent components of self-motion optical flow

Abstract

In this paper we describe the properties of independent components of optical flow of moving objects. Video sequences of objects seen by an observer moving at various angles, directions and distances are used to produce optical flow maps. These maps are then, recessed using independent component analysis, which yields filters that resemble the receptive fields of dorsal medial superior temporal cells of the primate brain. Contraction, expansion, rotation and translation receptive fields have been identified. Our results support Barlow's sensory coding theory and are in-line with other work on independent components of image and video intensities.