Abstract: A new feature-based image watermarking scheme robust to unclassified geometric attacks is proposed in this paper. Firstly, the multiscale Harris detector is utilized to extract steady feature points from the host image; then, the local feature regions (LFR) are ascertained adaptively according to the feature scale theory, and they are normalized by normalization technique; finally, according to the predistortion compensation theory, several copies of the digital watermark are embedded into the nonoverlapped normalized LFR by comparing the DFT mid-frequency magnitudes. Experimental results show that the proposed scheme is not only invisible, but also robust against common signals processing and unclassified geometric attacks.