Detail-preserving approach for impulse noise removal from images

An impulse removal algorithm is proposed to overcome the difficulty of detail preservation which is commonly encountered by existing methods. At the impulse detection stage, a directional impulse detection method is employed to avoid the misclassification of thin lines as impulses. As for the noise filtering stage, a hybrid algorithm which combines median-based filtering with pattern matching technique is adopted. Experiments show that the proposed approach outperforms a number of existing algorithms and the improvement on detail preservation is significant.