Denoising of document images based on similarity and connectivity principles

This paper proposes a highly efficient denoising technique in restoring document images that have been corrupted by impulse noise. The proposed algorithm consists of a number of shells and is based on the connectivity principle; whereby, pixels that have connections with at least one other pixel in each of the consecutive shells are deemed noise-free pixels. Simulation results indicate that the new denoising method is not only useful in eliminating the noise efficiently; it also keeps the fine details of the image.