Weight selection for the 2-d adaptive weighted median filter based on the gibbs random field model

The purpose of this paper is to introduce a new adaptive weighted median (awm) filter that preserves the details of different, textural areas within the image. In the development of the filter, a method is presented based on a Gibbs randoin field (GRF) model of the input and output images. It is shown that this method makes it possible to design and analyze filters with very different characteristics. The proposed awm filter is compared with the standard median filter and the linear averaging filter in noise elimination and detail preservation, and is shown to be significntly better in performance.