Bio-inspired disparity estimation system from energy neurons

Abstract

Bio-inspired control systems attempt to reproduce the intelligent behavior by simulating the internal architecture and mechanisms of the biological counterpart. In this paper we propose a bio-inspired algorithm for disparity estimation based on the disparity energy model. In literature several models have been proposed but actually each of these models seems to have unique features and unique lacks due to the intrinsic architecture. Different bio-inspired architectures are reviewed and from the comparison among these algorithms we propose a new architecture that takes into account the good properties of the previous models and tries to overcome the limitations. Simulations are performed on real images, and comparison with published algorithms on Middlebury stereo database are shown. We comment the obtained results from the point of view of humanoid robotics and show how the study of neurophysiology can inspire the design of vision systems.