Image Denoising with an Adaptively Weighted Four-Directional Total Variation Method

In this paper, an adaptively weighted four-directional total variation method (Ada-4WTV) is introduced to restore the image corrupted by additive Gaussian noise. In the Ada-4WTV, the trade-off parameter between regularization term and the data fidelity term is adaptive, which can help to deal with textures and smooth areas separately. What's more, four weights are designed in the TV regularization to preserve the textures and edges better. The boundary conditions are then also considered in the Ada-4WTV. Furthermore, the fast gradient projection (FGP) is extended to implement the Ada-4WTV. The experimental results indicate that the Ada-4WTV adapts to different noise levels and has better performance on keeping the image textures and edges.