According to the fact that the background of a small infrared target is sparse in the searching stage, a novel target locating method which uses a salient map based on the phase spectrum in an image is proposed. Compared with the traditional target search method based on sea-sky-line or coastline, this method has reduced its computational complexity greatly and overcome the shortage that the sea-sky-line or coastline can not be located accurately due to the unstable temperature. Compared with the method using an Itti model, the algorithm has overcome the weakness that the target can not be separated from the background. Two methods which use Discrete Fourier Transform (DFT) and Discrete Cosine Transform (DCT) to calculate the image phase spectrum are presented and their consistency is proved. A mathematic model which uses phase spectrum to calculate saliency map is established. Parameter selection of the model is explained. The theory and examples show that this algorithm is feasible and efficient.