Any motion of the camera during the image integration time may determine an image degradation known as motion blur. One approach to prevent this degradation, namely multi-frame image stabilization, consists of synthesizing a potentially motion blur free image by registering and fusing multiple short exposed image frames of the same scene. In this paper we propose an approach to image fusion for multi-frame image stabilization application. The proposed algorithm is robust to various disturbing factors that may occur in practice like: noise due to short frame exposures, blur in the individual frames, errors in the registration of the individual frames, as well as the presence of moving objects in the scene. We demonstrate the algorithm through a series of experiments and comparisons based on simulated test images as well as on real images captured with digital cameras.