Abstract: The fuzzy C means algorithm (FCM), widely used in the image clustering segmentation, has weak ability of resisting the noise. Although some modified FCM algorithms can reduce the effect of the salt and pepper noise as well as Gauss white noise, they can not effectively deal with the case that the image contains larger size noise. In this work, by combining with image filter technique, an FCM image segmentation algorithm is proposed. The size of filter window is determined by the size of noise, and the noise decision matrix is constructed by using the FCM clustering result and the noise decision rules are designed. In the algorithm, the filter operation in the filter windows is realized. The experiments are made on the image segmentation of the artificial gray image with noise and the fiber image, which show that the noise, especially the big size background noise, can be reduced effectively, and the satisfactory image segmentation results can be obtained.