Removal of Salt and Pepper Noise Using Sparse Representation

This paper proposes a new method to remove salt and pepper noise in gray scale images as well as color images. This is done in two stages. In the first stage impulse noise detection scheme is employed to detect the noisy pixel. This filter uses the morphological noise detector to classify the pixels as either corrupted or uncorrupted. In the second stage we treat the noisy pixel as a missing pixel in the image and apply in painting based on the sparse representations of the pixels that are identified as corrupted pixels in the first stage. Our experimental results show that the proposed method outperforms other best known denoising techniques used for the removal of salt and pepper noise, both visually and quantitatively. Extensive experiments have been carried out on different images to validate the efficiency of the proposed method.