Adaptive frame-based color image denoising

In this paper we study image denoising to restore color images contaminated by additive white Gaussian noise. For a color noisy image, block matching groups similar image patches together to form fourth order tensors. Using higher order singular value decomposition for tensors, we construct adaptive frames and propose an iterative thresholding algorithm to suppress additive white Gaussian noise in corrupted color images. The proposed algorithm can be viewed as an adaptive frame-based iterative hard thresholding method. We provide convergence analysis of the Nash equilibrium process raised in our proposed algorithm and any patch-based image denoising methods. Our experimental results show that the overall performance of our proposed algorithm is comparable to or often better than two well-known state-of-the-art color image denoising algorithms.