Memristor CNN Model for Image Denoising

In this paper we develop an algorithm for noise removal which extracts the target information (image) more precisely. The model is based on the convection direction during time evolution in order to diffuse the noise. For this purpose we propose a convection - diffusion system in which the convection term is added in the modified diffusion partial differential equation (PDE) and it serves as physical interpretation for removing the noise. We introduce memristor CNN (MCNN) model in order to make the numerical computations accurate. The proposed MCNN architecture has several advantages as high density, nonvolatility and programmability of synaptic weights. Finally, we provide simulations and validation on image denoising applications. In this way we show that our model is better in noise removing through several numerical experiments.