Blind Deblurring Based on an L_(1/2)/L_2 Regularization

JING Wen-feng,ZHAO Ren-xing,LI Zhi-min,SONG Wei,ZHENG Yan
DOI: https://doi.org/10.3969/j.issn.1005-3085.2015.05.004
2015-01-01
Abstract:Image deblurring is the basic work for image recognition and video analysis. In real-world applications, most image deblurring problems are ones of blind image deblurring.The problems are ill-posed and need to be solved by regularization methods. Since the existing regularization models for image deblurring are difficult to restore image details, we propose a novel blind deblurring model based on L1/2/L2 regularization and an alternating projection iteration algorithm to solve it. Experimental results demonstrate that the proposed model and algorithm have very good restoration on the detailed structure of original deblurred images,and have high computational efficiency and fine robustness to parameters as well.
What problem does this paper attempt to address?