Analysis and comparison of regularization techniques for image deblurring

Abstract

Image deblurring or deconvolution problems are referred as inverse problems which are usually ill-posed and are quite difficult to solve. These problems can be optimized by the use of some advanced statistical methods, i.e., regularizers. There is, however, a lack of comparisons between the advanced techniques developed so far in order to optimize the results. This paper focuses on the comparison of two algorithms, i.e., augmented Lagrangian method for total variation regularization (ALTV) and primal-dual projected gradient (PDPG) algorithm for Beltrami regularization. It is shown that primal-dual projected gradient Beltrami regularization technique is better in terms of superior image quality generation while taking relatively higher execution time. © Springer Science+Business Media Singapore 2016.