Rotational Tensor Nuclear Norm for Color Image Sparse Noise Removal

Abstract

Image Sparse Noise Removal constitutes a core issue within the domain of image processing. The tensor nuclear norm (TNN) has achieved great success in color image sparse denoising, but it suffers from the problem of direction sensitivity. In order to address this problem, a color image sparse noise removal scheme via the rotational tensor nuclear norm (RTNN) is proposed. First, to characterize the correlation across each dimension of the image, the RTNN is introduced. Then, on the basis of the RTNN, a model for image sparse noise removal is established. Next, an effective algorithm for the image sparse noise removal model is designed by adopting the ADMM. Finally, numerical experiments demonstrate that our scheme surpasses other compared methods with respect to evaluation metrics.