Novel Image Deconvolution Algorithm Based on the ROF Model

Abstract

The ROF (Rudin-Osher-Fatemi) model is one of the most successful and widely used models in image deconvolution. However, efficiently solving this model is usually impractical because of its non-smoothness. To solve the ROF model quickly and accurately to improve image deconvolution, a novel algorithm based on the splitting Bregman method and the two-step iterative thresholding method (2-SITM) is presented. The ROF model is decomposed into several sub-problems using the split Bregman method. These sub-problems are then solved by corresponding methods to obtain their closed-form solutions. To compute the denoising sub-problem, the 2-SITM is introduced. Compared with the popular iterative thresholding method, the 2-SITM is conducive to improving the performance of the presented algorithm. In an experiment, uniform-blurry images are restored by the 2-SITM to verify its effectiveness. Results also show the superior performance of the presented algorithm to some similar state-of-the-art algorithms.