A new bio-inspired algorithm for early vision edge detection and image segmentation

Abstract

We present a new bio-inspired algorithm for early vision image preprocessing that includes both edge detection and image segmentation. Previous research efforts either implemented Difference-of-two-Gaussians (DoG) bandpass filtering with resistive networks for edge detection or image segmentation with so-called resistive fuses. By using a feedback mechanism, we combine the superior edge detection algorithm of two-layer parallel resistive networks with image segmentation. To allow an early estimation of the results achievable with this algorithm, a special simulation program was developed which simulates the behavior of an idealized electric circuit implementation of this algorithm. Simulation results not only for the new proposed algorithm but also for two-layer resistive networks and for networks with resistive fuses are presented. Moreover, the development of the simulation program also led to a new effective method for subsequent implementation of the algorithm.