Defocus deblurring with a coded aperture

Abstract

A conventional camera always captures blurred images of scene information away from the focal plane. Deblurring those blurred images is a very important and challenging job in many fields including optics, astronomy, computer vision and computer graphics. Since the defocus kernel plays a key role in the image degradation and the image recovery, many people have bent themselves to the study on the defocus kernel and achieved a lot of results. In this paper, we insert a broadband mask into the aperture of the camera lens to preserve more high frequency information for facilitating the defocus deblurring. We propose a genetic algorithm for computing the defocus kernel of a fixed distance away from the focal plane using a given sharp image and a defocused one. Finally, we implement extensive experiments to show that the computed defocus kernel can produce better images.