A method for small IR target detection in complex background is proposed in this paper. Based on the multiresolution decomposition of Orthogonal Wavelet Transform (OWT) ,the important information of the target or background can be obtained , and the noise and clutter can be suppressed by selection of frequency. To detect the horizontal edge of the large scale approximate , the sea2sky2line can be located while the target potential area can be decided. The fusion algorithm is used to the horizontal and vertical directional details and an adaptive threshold is calculated to binary the image. By searching the target potential area and assigning threshold , the false alarm points are discarded and then the target can be segmented. It is proved experimentally that this algorithm is efficient to detect small IR target in a sea surface background.