IMPULSE NOISE REDUCTION BASED ON SIMILARITY AND CONNECTIVITY OF PIXELS

Abstract

In this paper, we propose an effective noise reduction method for image corrupted by impulse noise in highly noise ratio. Preserving edges and details in the process of impulsive noise filtering is an important issue. Image quality is determined by human eye. As a result, the basic strategy of the proposed filter is to exploit the similarity and connectivity between the central pixel and its neighbors. We calculate the number of these pixels for judging whether the central pixel is corrupted by noise or not. Then, we use the local image features to estimate the original pixel value of the noisy pixel. In our experimental results, we shall show the proposed noise reduction method is better than other method in highly corrupted images.