Discrete Total Variation Model with Gradient Fidelity Term for Image Restoration

In this paper, we introduce a new discrete model for image denoising. We first smooth the gradient field of the observed image by a discrete total variation model. Then we construct a new discrete functional with the smoothed gradient fidelity term, which can alleviate the staircasing effect efficiently and preserve sharp discontinuities during the images denoising. Here, the difference discrete variation principle is used to get the discrete Euler-Lagrange equation. We also discuss some numerical experiments which prove our proposed model and algorithms to be more efficient.