A new filtering algorithm with the purpose of eliminating impulsive noise is presented. In order to keep the content of signals as much as possible when the noises are removed, the following approaches are developed in this paper. Noisy items are identified by inspection of their surrounding neighborhood, and afterwards their values are replaced with the most "conservative" ones out of their neighbors' value. If a pixel value is the extremum (max. or min.) of its neighborhood, it is a noise pixel, or else it is a signal pixel. Experiment results show that the new algorithm is good in performance to keep the content of image and remove noise, especially in the case of strong noise.