Research And Implementation Of Motion Deblurring Algorithm Based On Kernel Estimation

A single image deblurring is an important task in the image restoration. Due to image perform perspective motion in the process of transimission. Which makes the image there is some degree of fuzzy, although there are many image deblurring method. But most methods are heavy computation, slow speed, small kernel assumption, ringing artifact. In the process of kernel refinement, achieved iterative support detection (ISD) algorithm,For shift-invariant motion blur. Algorithm is realized in this paper, in the phase of blur kernel initialization has carried on the edge image prediction. So choicing the image gradient effectively and processing the large size of blur kernel. This algorithm enforce the sparse of blur kernel, so as to improve the computing speed. While in the process of image deconvolution restoration. We employ the TV-L1deconvolution model for suppressing noise.Experiment results show that, the achieved motion deblurring method has good performance in precision and computation speed. The method of motion deblurring shows robustness to both simulate and actual license plate image.