Abstract: Against most of the existing watermarking algorithms for remote sensing image are based on global image, their resistance to rotation, scaling, ranks of shear and combined geometric attack is poor. It proposes a new watermarking scheme for remote sensing image based on Sift feature region and Contourlet transformation. It selects a binary digital image as watermark image, after scrambling the watermark image it scans the scrambled image into one-dimensional vector and then uses the binary sequence as watermark information to be embedded; secondly, it extracts Sift feature points in the host remote sensing image, filters the feature points and determines feature regions for watermark embedding; finally, the selected feature region is specified and normalized, it is applied Contourlet transformation and watermark information is embeded into the selected transform coefficients. Experiments have shown that the algorithm has good robustness against conventional signal processing such as JPEG compression, noise attacking, filtering, and etc., at the same time it better resists to geometric attacks such as rotation, scaling, moving and their combinations.