A novel image segmentation method was proposed in view of the features of forward-looking infrared images used for implementing the image segmentation under complex background. The method adopts the technique of background suppression, mutual information and mathematical morphology. First, the background was suppressed by using median filter. Then, mutual information was used to determine the optimal threshold, which applies the method of image registration to image segmentation. Finally, the threshold image was filtered by morphological filter and then accurate target region was obtained. The experiment results show that the method performs well in target segmentation under complex background. The segmentation results of the method are of great help to the target recognition and tracking in the next step.