Image Reduction is one of the most important image processing tasks. In this article, a new fuzzy filter is presented for the removal of impulse noise in gray and color images. The filter consists of two stages. In the first stage, using trapezoid membership function four values are used to detect whether a given pixel contains noise or not. In the second stage fuzzy enhanced classifier is applied successively to remove the impulse noise in gray and color images. Experimental results are obtained to show the feasibility of the proposed approach. These results are also compared to other filters by numerical measures and visual inspection.