Multisensor data fusion has played an important role in image processing recently. For some images from the same scene

each of them has different noise because of their different sources. This paper presents a new kind of image data fusion algorithm based on the self organizing feature map neural network. This algorithm can be performed with three steps. In the first step

the pretreatment of the images is performed by the weighted median filter in order to remove some noise. In the second stage we use self organizing feature map neural network to cluster the pixels of each image and then extend hard partition into fuzzy partition. In the third stage

we fuse the data from the last step in conformity to a certain rule. The simulation results illustrate that this new algorithm can improve the quality of the image distinctly and the pretreatment of the images can improve the fusion result efficiently.