Rotation, scaling, and translation invariant image watermarking based on radon transform

Abstract

Geometric distortions are simple and effective attack to many existing watermarking methods. We propose a rotation, scaling and translation (RST) resilient watermarking method through embedding watermark in RST invariant derived from Radon transform and Fourier transform. Based on the translation and rotation properties of Radon transform and the translation invariant property of Fourier magnitude, the RST invariant is obtained. The watermark is embedded in middle frequency of RST invariant domain, and the watermarked image is obtained with the inverse procedure. In detection process, the image which need to be detected is first transformed to the translation and rotation invariant domain, the normalized correlation between the watermarked coefficients and watermark is calculated. The correlation is compared with a threshold in order to determine if the watermark is present. To resist scale attack, the scale factor need be estimated by exhaustive search. The experiment results show the scheme is robust to RST attack and JPEG compression.