Weighted structure tensor total variation for image denoising

Based on the variational framework of the image denoising problem, we introduce a novel image denoising regularizer that combines anisotropic total variation model (ATV) and structure tensor total variation model (STV) in this paper. The model can effectively capture the first-order information of the image and maintain local features during the denoising process by applying the matrix weighting operator proposed in the ATV model to the patch-based Jacobian matrix in the STV model. Denoising experiments on grayscale and RGB color images demonstrate that the suggested model can produce better restoration quality in comparison to other well-known methods based on total-variation-based models and the STV model.