Efficient denoising of images using a nonaggressive median filtering scheme

The paper proposes a nonaggressive median filtering scheme, with two settings, that can be used to restore images moderately corrupted with either of the two distinct types of impulse noise-fixed-valued and random-valued. The scheme benefits from the fact that its design requires neither the difficult selection of optimal image-data dependent threshold(s), nor any kind of prior training. The filter has been tested on various standard images contaminated with both varieties of impulse noise (not added simultaneously); the results obtained strongly validate the good performance of the process. The proposed approach fulfils all the requirements needed for use as a real-time video-processing component.