Motion Blurred Images Restoration Based on Sparse Regularization

Lei Li,Jiang Ming Kan,Yexiao Wu,Wenbin Li
2013-01-01
Abstract:Motion blurred images restoration is an unavoidable problem in computer vision application area. A method of motion blurred images restoration based on sparse regularization is proposed in this paper. Original image, constructed original image with sparse characteristic of motion fuzzy kernel of wavelet frame transform and regularization term of motion fuzzy kernel are used in this method. The blind deconvolution problem of motion blurred image restoration is simplified to two non-blind deconvolution minimum problems with this method, and Bregman iteration method is finally adopted to solve these two minimum problems. Experiments are conducted on blurred image with uniform linear motion, uniform curvilinear motion and arbitrary curvilinear motion. It shows that the satisfactory result is achieved with the proposed method, and its structural similarity index is better than the algorithm proposed by Fergus and Shan.
What problem does this paper attempt to address?