Fast impulsive noise removal in color images

In the paper a novel switching filter based on the reduced ordering statistics, combined with the peer group concept is proposed. The new technique assigns to each pixel the sum of the distances to two closest neighbors and if the sum exceeds a specified threshold, then the pixel is declared as corrupted and replaced, otherwise it is retained. The main advantage of the novel approach is its ability to suppress the noise component, while preserving fine image details. The proposed filtering framework is characterized by a very low computational complexity, which enables its adoption in real-time applications.