Nature-inspired computer vision algorithms with reaction-diffusion models describing pattern formation processes

点击翻译按钮获取中文摘要

A reaction-diffusion model with an excitation-inhibition mechanism describes pattern formation processes observed in nature. In this research work, we built computer vision algorithms by utilizing reaction-diffusion models. In particular, we focused on a reaction-diffusion stereo algorithm, and proposed to integrate image edge information into the algorithm. We found that the imposition of strong inhibitory diffusion on the algorithm at image edge positions is effective with respect to accuracy of stereo disparity detection for several test stereo images.