Random-valued impulse denoising using sparse regularizations

In this paper, we propose a novel two-stage algorithm for random-valued impulse denoising based on sparse representations and image inpainting. The overall algorithm consists of a detection stage followed by a filtering stage. In particular, first impulse locations are detected by applying sparse denoising and grey relational analysis. In the second stage, the detected impulses are replaced with reconstructed pixels obtained from an image inpainting method. Extensive simulation results show that the proposed method is effective in removing random-valued impulse noise from highly textured images at high noise ratios compared to some of the existing techniques.