Abstract
Abstract. Image denoising is one of the important tasks required by medical imaging analysis. In this work, we investigate an adaptive variation model for medical images restoration. In the proposed model, we have used the first-order total variation combined with Laplacian regularizer to eliminate the staircase effect in the first-order TV model while preserve edges of object in the piecewise constant image. We also propose an instance of Split Bregman method to solve the proposed denoising model as an optimization problem. Experimental results from mixed Poisson-Gaussian noise are given to demonstrate that our proposed approach outperforms the related methods.
Highlights
In the process of receiving and transmitting through communication channels, the image is usually corrupted by noise of a different nature
Many types of noise have been studied in the literature, but the majority of models in image denoising are associated with additive Gaussian noise and Poisson noise (Bertero et al, 2009, Pham, 2017, Pham, 2018)
Authors in (Pham et al, 2018) proposed a modified scheme of gradient descent (MSGD) that is impossible to avoid sign-changing of the solution during the optimization process and guarantees the reconstructed image to be positive in the image domain
Summary
In the process of receiving and transmitting through communication channels, the image is usually corrupted by noise of a different nature. Problems of negative values arise in the numerical algorithms To avoid this problem, authors in (Pham et al, 2018) proposed a modified scheme of gradient descent (MSGD) that is impossible to avoid sign-changing of the solution during the optimization process and guarantees the reconstructed image to be positive in the image domain. No previous training is required (by contrast with deep learning techniques), and only one observation of the image is needed This method is slow due to the constraint of stability conditions about the time step size. The model (1) performs well in preserving image edges compared with other related method This model has sometimes undesirable staircase effect in some cases, namely, the transformation of smooth regions into piecewise constants regions (Chan et al, 2015). Experimental results show the effectiveness of the proposed method for denoising medical images corrupted by mixed Poisson-Gaussian noise. Comparison with other related methods and state-of-the-art algorithms is provided numerically as well
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
