A New Blind Deblurring Method via Hyper-laplacian Prior
Jun Kong, Kesai Lu, Min Jiang · Procedia Computer Science · 2017
Blind image deblurring is always a thorny problem due to its uncertainty. In this manuscript, we proposed a new blind deblurring method in the maximum a posterior (MAP) framework to remove the blur of the image. Firstly, we choose the hyper-Laplacian prior to be a regularization of the gradients of an image. Secondly, we adopt an operator called generalized soft thresholding (GST) to solve the non-convex problem during the whole deblurring process. Thirdly, we compared our method with some popular approaches in both quantitative and qualitative aspects. Experimental results show that the method proposed by us performs much better than the other approaches.