Robust Image-Processing Filters Preserving Small Features and Rejecting Impulse Noise

New robust image-filtering algorithms based on RM point estimates and on the known KNN-filter technology are presented. These RM-KNN filters show a sufficiently high efficiency in rejecting pulse noise and preserving object boundaries and small features of the image. The filters were checked with test and real images typical of remote-sensing problems. The proposed filters provide a good visual quality of the filtered images and feature better characteristics than the standard median filter.