A Nonlocal Denoising Framework Based on Tensor Low Rank Representation with Weighted Tensor Schatten p norm

Image denoising plays an indispensable role in image processing, so how to more effectively solve this problem has become an important research topic at present. The tensor low rank representation method based on global low rank has achieved great success in image denoising. However, these methods ignore the nonlocal self-similarity of images, which exists in a wide range of multi-dimensional data such as video sequences. In order to integrate the global low rank and nonlocal characteristics of images, this paper proposes a denoising model that combines nonlocal self-similarity and tensor low rank representation. Specifically, similar patches are grouped into a tensor according to nonlocal priors, the tensor is denoised, and then the denoised tensor is summarized to obtain the denoised image, which can make full use of the structural redundancy of the image. The results in image and video denoising experiments validate the effectiveness of our method and indicate that it outperforms other comparison methods based on nonlocal characteristics.