For the purpose of improving high-resolution remote sensing images registration precision,nonsubsampled contourlet transform is applied to high-resolution remote sensing images registration. Firstly,for nonsubsampled contourlet transform shift invariance,nonsubsampled contourlet transform is used to extract the images edge in contourlet transform domain and the feature points are extracted from images edge by selecting an appropriate threshold. Then,normalized cross-correlation matching and probability support matching method are used to match the images feature points. Finally,triangle-based local transformation function is employed to register the images. The experimental results show that this method can more accurately extract the corresponding feature points of high-resolution remote sensing images and increase correct matching probability and have more precise and more robust registration effect than the method based on wavelet.