Abstract: In order to improve the robustness of the digital watermarking image against common image processing and complex geometric attacks, a novel robust watermarking scheme of digital image was proposed by using the stable SIFT(scale invariant feature transform) feature points and the cross-ratio values with the characteristic of invariability. Firstly, the original digital image was decomposed with NSCT (non-subsampled contourlet transform), thus the SIFT feature points were extracted in the NSCT low-frequency coefficients. By optimizing and screening these SIFT feature points, some stable SIFT feature points were gotten and some quadrilateral regions which met the needs of “near-convexity square”were searched out in these stable feature points. These quadrilateral regions would be used as local watermarking regions. Secondly, some cross-ratio values designed in advance were used to determine the low-frequency coefficients to be embedded watermarking information. Lastly, the watermarking information was embedded in these low-frequency coefficients by using odd-even quantization index modulation method. Compared with the traditional algorithms, the image embedded watermarking information and attacked by geometric distortion does not conduct inverse geometric transform when the watermarking information was extracted. Experimental results show that the proposed scheme has a good ability of robustness against common image processing and various geometric attacks.