Because of many factors, the infrared star images acquired in the daytime have a low signal-to-noise ratio and a nonuniform background. The star targets in the images can not be extracted by using common filtering methods. Currently, the common morphological method is also not ideal for the processing of star images. Firstly, the background characteristics are analyzed and the multi-frame superimposition method is used to reduce the random noise, so as to improve the signal-to-noise ratio. Then, setting a threshold, the background in the superimposed star image is removed and the image only containing targets and noise is obtained. Finally, the star image is processed by using a template similar to the star in size according to certain rules so as to separate the targets. The experimental result shows that this method is more efficient in star target separation.