Image Restoration for Linear Local Motion-Blur Based on Cepstrum

Abstract

This paper presents a real-time restoration method for linear local motion-blur image. for an image in which only the fast moving-object is blurred but the background is clear, the proposed basic strategy is to divide such an image into many sub images firstly and then detect the blurred sub image by the gradient distribution and the maximum of cepstrum. for a blurred sub image, the blur direction and blur length are estimated to calculate the parameters of point spread function (PSF) and Lucy-Richardson deconvolution algorithm is employed to restore this blurred sub image. Using many artificial and real blurred images, the experimental results show that the proposed approach is more accurate and robust than other methods.