Implicit method for nonlinear complex diffusion with applications to image denoising

In this paper we focus on the development and implementation of an implicit finite difference method for solving a complex diffusion differential equation with applications to noise filtering in images. We give a complete characterization of the implementation of the fully implicit method, including the characterization of the derivatives for Newton's method at each step. In order to show the feasibility of the method, we compare the implicit method with the semi-implicit and apply it to both in vivo images of the human retina by optical coherence tomography (OCT) and standard images (Lena and the Cameraman). We compute noise removal performance metrics to show the effect of several parameters in the performance of the method.