A rotation, scaling, and translation invariant local watermarking is proposed with one or two Krawtchouk moment(s) of the original to estimate the geometric distortion parameters including rotation angle, scaling factor, and translation parameter. Krawtchouk moments can be used as private key of watermark extractor. Watermark is inserted into perceptually significant Krawtchouk moments of original, and watermark based on Krawtchouk moments is local. Independent component analysis (ICA) is utilized to extract watermark blindly. Experimental results show that this method has a good robustness against distortions preformed by watermark benchmark -- Stirmark.