In the infrared imaging and tracking system, infrared grey image of small targets is disturbed by the background and noise, that results in low SNR. The grey information of the target is required in the post-processing of target recognition. The mathematical morphology based on algorithm to detect small targets is designed for infrared grey image and implemented by field programmable gate array. It is shown that mathematical morphology filtering can significantly enhanc SNR of the target. In the filtering the disturbance is eliminated, and the grey information is remained. The algorithm is adaptable to real-time image processing. It is effective for detecting small targets in infrared grey image.