Bridge over water is a typical man-made target. Using the high-resolution optical remote sensing images, to extract bridge over water appears significant for civilian, military, and commerce. First, in the present paper, according to the spectral characteristics of water in near-infrared spectrum, water information is extracted using the iterative method for the threshold value, to limit the spatial extent of bridge extraction. Second, mathematical morphology is performed on the water information, to connect the separated water bodies due to the presence of bridges on the image. Third, overlay analysis is conducted between the two images before and after mathematical morphology operations, to extract the potential bridge. Finally, based on priori-knowledge of the bridge, the false bridge is removed. In this paper an experimental area is selected to verify the effectiveness and applicability of the method. Experimental results show that the method is effective for bridge extraction over water by using high-resolution remote sensing images based on spectral characteristics of ground objects.