Digital watermarking robust to geometric distortions

Abstract

In this paper, we propose a digital watermarking scheme that is robust against geometric distortions. The proposed method uses image moment normalization and a correlation peak position modulation (CPPM) to recover geometric distortions. This mechanism is invariant to affine transform attacks and suitable for public watermarking applications, where the original image is not available for watermark extraction. In addition, CPPM ensures that there is sufficient capacity to hide specific information. We verify the robustness of our approach to rotation, scaling, and Gaussian noise attacks. The watermark can be detected with zero BER, because disadvantages disappear when CPPM is combined with image moment normalization-only the advantages of the two methods remain. © 2012 Springer-Verlag Berlin Heidelberg.