Image Denoising via Robust Simultaneous Sparse Coding

Simultaneous sparse coding (SSC) has shown great potential in image denoising, because it exploits dependencies of patches in nature images. However, imposing joint sparsity might neglect the sight difference between patches. In this paper, we propose an image denoising algorithm based on robust simultaneous sparse coding (RSSC). In our algorithm, the sparse coefficient matrix is decomposed into two parts. One coefficient matrix is imposed on the joint sparse regularizer which exploits self-similarities of image patches while the other matrix is imposed by the elementwise sparse regularizer which considers the subtle differences between patches. Experiments on the benchmark data show the superior performance over the state-of-art algorithms.