Reaction-Diffusion Systems in Pattern Formation and Pattern Recognition Processes

This paper presents a quick review of reaction-diffusion systems that self-organize spatio-temporal patterns in chemical and biological systems. A pair of reaction-diffusion equations having activator and inhibitor variables is a typical model of equations describing their pattern formation processes. Some of the reaction-diffusion systems can realize several functions of image processing such as edge detection and segmentation required for pattern recognition processes. Thus, reaction-diffusion systems which have been studied as pattern formation processes are also interesting topics as pattern recognition processes. The authors have recently done the research of realizing functions of edge detection, segmentation and stereo disparity detection by utilizing reaction-diffusion equations with large inhibitory diffusion. This paper additionally presents a brief introduction of our recent research topic, in particular, an edge detection algorithm with a discretely spaced system of the FitzHugh-Nagumo reaction-diffusion equations. Experimental results obtained for artificial and real images show the comparison between a previous standard algorithm and the reaction-diffusion algorithm designed for edge detection. Key–Words: Reaction-diffusion, pattern formation, pattern recognition, long-range inhibition, image processing, computer vision, edge detection, non-linear reaction