Two-stage Image Denoising Based on Sparse Representations

It remains a challenging task to restore the image which is contaminated with heavy noise. In this paper, an image denoising method is proposed based on sparse representations. In the dictionary training stage, a matching criterion based on correlation coefficient is introduced and a dictionary pruning scheme is proposed to tackle the conflicting issues of structure extraction and artifact suppression. Experimental results show that the proposed method achieves significant improvements over the previous sparse denoising methods and outperforms the state-of-the-art methods in terms of both objective and subjective quality at high noise level.