Segmentation and Edge Detection of Noisy Image and Low Contrast Image Based on a Reaction-Diffusion Model

Abstract

An increasing attention is focused on information processing by reaction-diffusion system, in which temporal and spatial patterns are self-organized. In the system, two interesting phenomena of Turing pattern formation and stochastic resonance have been reported. We have been proposed a new approach for image segmentation and edge detection based on a reaction-diffusion model (Fitz-Hugh & Nagumo (FHN) model). In this paper, noisy image or low contrast image are tested to confirm effectiveness of the method. Compared to the conventional method, the Turing condition realizes more reliable tool for noisy image segmentation. And, addition of moderate noise improves the performance of image segmentation. Stochastic resonance condition acts as more powerful tool for edge detection and segmentation for low contrast image.