This paper presents an ordering threshold switching median filter to solve the contradiction between noise attenuation and image detail preserving. From the ordering information of the pixels in the window

and based on extremum median filtering the image corrupted by impulse noise is divided into three pixel classes

that is

noise pixels

edges and details

and smooth regions. With the statistic of a lot of standard images tested

the parameters of the classifier are properly chosen in order to deal with most images adaptively. Then switching median filtering is applied with the classifier. Therefore the smooth regions and noise pixels are filtered by median filters that have a good noise removing capability

especially with the 'salt and pepper' noise. However

most of the edges and details of the image are untouched

so that the restored image can keep details even in variable magnitude impulse noise conditions. A comparison of median filter

extremum median filter and the method in this paper is provided both in subjective images and objective MAE and MSE data. Obviously

the results indicated that the new method has better properties.