期刊名称：International Journal of Signal Processing, Image Processing and Pattern Recognition
印刷版ISSN：2005-4254
出版年度：2016
卷号：9
期号：5
页码：389-398
DOI：10.14257/ijsip.2016.9.5.35
出版社：SERSC
摘要：This paper models image deconvolution as an l 2 -l 1 minimization problem, which is an approach taken by many state-of-the-art image deconvolution algorithms. We present a novel iterative algorithm based on the split Bregman method and the stationary second- degree method, which efficiently addresses the classic convex minimization problem. The split Bregman method, which has been proven to be very efficient for non-differentiable minimization problems, decomposes the equivalent constrained version of the l 2 -l 1 deconvolution problem into a series of sub-problems. These sub-problems are then individually solved using appropriate methods to obtain their closed-form solutions. Unlike the majority of other similar deconvolution algorithms, we use a modified stationary second-degree method to solve the l 2 -l 1 denoising sub-problem, prompted by some recent work on the improvement of the iterative thresholding method. The presented algorithm can be categorized as a split Bregman method, so convergence of the solution can be guaranteed. In our experiment, the presented algorithm and the algorithms in references [6] and [8] are used to restore Gaussian-blurry and uniform-blurry images. The experimental results show that the presented algorithm is effective and it outperforms other algorithms in comparison.
