The proposed approach adopts adaptive filter to suppress random noise and Gauss's noise so as to enhance the SNR. Then, Otus's threshold method is used to segment the image and locate the sea-sky-line, through which the target potential area can be determinated. A mathematical morphology filter, Top-Hat operator, is employed to execute filtering to restraint the background contribution and maintain the small target in high brightness area. By choosing proper structural element, an open filter is executed to discard small false alarms. Eventually, the real small target can be segmented through searching maximal value in the target potential area nearby the sea-sky-line and assigning threshold. The experimental results show that the method can reject the clutter caused by waves and clouds and effectively detect and segment IR small target in the complicated sea background.