An algorithm for restoring fisheye image using nonlinear inverse diffusion equation

摘要

When we are collecting traffic information from CCTV images, we have to install a detection zone in the image area while the pan-tilt system is on duty. Automation of an installing detection zone with pan-tilt system is not easy because of machine errors. Therefore, a fisheye lens camera or convex mirror camera is necessary for getting wide area images. There is a problem, in this situation, decreased system speed or image distortion. The distortion is caused by occlusion of angled rays as in snapshots taken by a trembling person. In order to overcome distortion, in this paper, we propose a speedy method of de-blurring by nonlinear inverse diffusion equation. The result of applying de-blurring methods shows that the proposed method has 11% of increased MSE more than Landweber method.