Preprocessing in a reaction-diffusion algorithm designed for image edge detection

This paper proposes a reaction-diffusion algorithm for image edge detection in the framework of FitzHugh-Nagumo model. FitzHugh-Nagumo model has two variables, activator and inhibitor, which are governed by two time-evolving differential equations, respectively, for simulating a process of biological excitation and inhibition phenomenon observed along a nerve. The proposed algorithm places FitzHugh-Nagumo elements, which contains a pair of activator and inhibitor variables, at the image grids. At first, the algorithm gives initial conditions of the elements according to an inputted gray level image. Then, it performs preprocessing for reducing noise by using only inhibition equation at the elements, and finally performs edge-detection by using both excitation and inhibition equations. The performance of the proposed algorithm is investigated with artificial and real images.