Box constrained total generalized variation model and primal-dual algorithm for Poisson noise removal (Q2193940)

In this paper, variational based approach is used for image denoising. The authors propose a method for Poisson noise removal from the 8-bit gray scale images. The method is based on a box constrained total generalized variation (TGV) model that extends the total variation (TG) one introduced by \textit{L. I. Rudin} et al. [Physica D 60, No. 1--4, 259--268 (1992; Zbl 0780.49028)]. The model is further reformulated as a minimax problem and is solved by the Chambolle-Pock's first-order primal-dual algorithm. The saddle point of the minimax problem is computed alternately with the primal variables and the dual variables fixed alternatively. Experiments with the proposed denoising model were performed on standard benchmark images corrupted by various Poisson noise intensities. The results of experiments demonstrate that the proposed model both gets better visual effects and also better images quality measures (signal to-noise ratio, peak signal-to-noise ratio and structural similarity index) than several existing state-of-the-art methods. The proposed model seems to be effective in eliminating the staircase effect and avoiding the blurring edge.