The lo-norm-based Blind Image Deconvolution: Comparison and Inspiration
Single image blind deblurring has been intensively studied since Fergus et al.s variational Bayes method in 2006. It is now commonly believed that the blur-kernel estimation accuracy is highly dependent on the pursed salient edge information from the blurred image, which stimulates numerous l0-approximating blind deblurring methods via kinds of techniques and tricks. This paper, however, focuses on the four recent daring attempts which are all based on the simple and direct lo-norm. A systematic com- parative analysis is made towards those methods, clarifying their similarities and differences, and providing a benchmark evaluation on both the deblurring quality and computational efficiency. Results have demonstrated that the lo-norm alone is far enough to achieve top blind deblurring performance. Instead, details are to be paid with fairly more attention as working on the problem formulation as well as the algorithmic deduction. Inspired by the success of the bi-lo-l2-norm regularization, an attempt has been made to boost a recently proposed normalized sparsity-based blind deblurring method via simply borrowing core ideas behind the bi-lo-l2-norm regularization. Experimental results show that the boosting approach has leaded to a significant improvement in terms of both accuracy and efficiency. Finally, several possible extensions are discussed towards the bi-lo-l2-norm regularization.
Keywords: blind deblurring; camera shake removal; variational Bayes; lo-norm minimization; split Bregman; half-quadratic
ABOUT THE AUTHORS
Hai-Song Deng was born in Xuzhou, Jiangsu Province, China. She received the B.S. degree in Mathematics in 2003 from Xuzhou Normal University, Xuzhou, China, the M.S. degree in Statistics in 2006, and the Ph.D. degree in Management Science and Engineering in 2011 both from NUST, Nanjing, China. Since 2011, she served as an assistant professor at School of Science in Nanjing Audit University. Her research fields include Bayesian variable selection, sparse optimization, surrogate modeling, design and analysis of computer experiments, and so on.
Wen-Ze Shao was born in Ganyu, Jiangsu Province, China. He received the B.S. degree in Science of Information and Computation in June 2003, and the Ph.D. degree in Pattern Recognition and Intelligent System in July 2008, both from Nanjing University of Science and Technology (NUST), Nanjing, China. From June 2003 to December 2011, he served as an officer in the People's Liberation Army (PLA) of China. In January 2012, he joined Nanjing University of Posts and Telecommunications (NUPT) as an assistant professor. Since May 2014, he also worked as an academic visitor (Postdoc) at Department of Computer Science in Technion-Israel Institute of Technology. He is particularly interested in the fields of natural image modeling, synthesis and analysis sparse modeling, statistical machine learning, variational PDE and Bayesian methods with applications to innovative and interactive imaging and vision problems.