Salt-and-pepper noise removal based on nonlocal mean filter

In this paper, a new two-phase method for salt-and-pepper noise removal is proposed which combines the adaptive median filter and nonlocal mean filter. In the first phase, the adaptive median filter is used to identity pixels which are likely to be contaminated by noise. In the second phase, the image is restored using the nonlocal mean filter which is firstly proposed for Gaussian noise removal. It has a strong ability to handle textures and repetitive structures. Experimental results show that the proposed algorithm achieved not only high PSNR but also pleasure visual results even when the noise level is high as 90%.