Unrolling Multi-channel Weighted Nuclear Norm Minimization for Image Denoising

Abstract

We propose an unrolled deep network that integrates the flexibility of model-based algorithms and the advantages of learning-based algorithms. Specifically, based on the multi-channel optimization model for real color image denoising under the weighted nuclear norm minimization formulation, we propose an algorithm for image denoising that can learn the weights for nuclear norm from training datasets through end-to-end training. Experimental results show that the proposed algorithm achieves better performance than traditional iterative algorithms.