Iterative shrinkage thresholding algorithm with redundant dictionary for image denoising

Upon the state-of-art technique of image sparse reconstruction, a new image denoising algorithm based on l₁-norm model is proposed in this paper. Without using the common transform bases, firstly the redundant dictionary trained by K-SVD algorithm is used as sparse representation for different image models. Then followed by the denoising algorithm consist of fast iterative shrinkage thresholding algorithm and least squares solution. The simulation results for both the current l₀-norm model based method and the proposed method demonstrate our method is more robust than the current method in terms of the peak signal-to-noise ratio.