Frequency domain analysis can well detect the edge of the salient region in the remote sensing imagery detecting. But it may mistakenly regard the inner parts of the saliency region as the background. A new algorithm based on multi-scale fusion techniques of the image high frequency information is proposed. First, the new algorithm creates several spatial scales of remote sensing images by using Gaussian pyramid. Then, for each scale, the new algorithm can get the high frequency information by the Fourier transform. Finally, the new algorithm gets the final saliency map by fusing the high frequency information on one scale. The new algorithm can not only well extract details of the salient region, but also effectively get rid of mistaken detection of the inner parts of the saliency region. Comparing with Itti model, the new algorithm has lower computation complexity.