The median filters have been generally adopted to reduce impulse noise.Although many existing median approaches have been greatly improved

the results of image filtering were not satisfactory at high noise density.Thus

in this paper

a novel median algorithm is proposed.First

all pixels of the filtering window are classified.Then the median subset of the multi-elements subsets is educed.The number of the multi-elements subsets determines whether the image is filtered or whether the filtering window is extended.Finally

in simulations

the proposed algorithm is compared with several kinds of median filters in the numerical value and the vision

respectively.Experimental results show that the new filter can effectively reduce the impulse noise and preserve more details of original images.