Images captured by the staring imaging system in geostationary orbit usually emerge blur because of the satellite vibration in the exposure time. To eliminate the motion blur degradation, the paper proposes to capture multiple short-exposure images and fuse them on sub-pixel scale instead. Image sequences with different conditions are captured by a high-speed camera, cooperated with a vibration simulation system. The sub-pixel displacements between the different frames estimated by image registration algorithm are compared with the motion data measured by the linear encoder. Experimental results indicate that the algorithm precision is better than 0.1 pixel. Several sharpest short-exposure frames are chose for sub-pixel image fusion. Compared with the long-exposure image, the motion blur in the fusion image is weakened. It proves that the method of multiple integration and sub-pixel image fusion is powerful of vibration suppression for staring imaging system in geostationary orbit.