Night-time Pedestrian Detection by Visual-Infrared Video Fusion
Yuxi Chen, Chongzhao Han · 2008
A method of night-time pedestrian detection by fusion of visual and infrared video was proposed. At first, infrared video is denoised by spatio-temporal filtering, then bright regions which corresponding to high temperature regions in infrared video frames are segmented by seeded region growing algorithm, then pedestrian are detected by fusion of region-color and region-shape information, at last, infrared video and visual video are fused to enhance detection result. Results show that combination of CCD and IR camera sensors can help to solve night-time surveillance problem.