A Dft-based Semi-blind Multiple Watermarking Scheme for Images

DWT and DCT domain watermarking schemes are not robust to geometric attacks such as rotation, scaling, and translation. Recent research has therefore focused on DFT-based watermarking because of its properties. In this paper, we generalize a circular watermarking scheme by embedding one watermark in lower frequencies, and another in higher frequencies. Our experiments show that if the circular watermark is embedded in higher frequencies, the percentage of false negatives is higher for one group of attacks (i.e, JPEG, Gaussian noise, blurring, and resizing), and lower for another group of attacks (i.e., histogram equalization, contrast adjustment, gamma correction, scaling, rotation, and cropping). Similarly, if the circular watermark is embedded in lower frequencies, the percentage of false negatives is lower for one group of attacks (i.e, JPEG, Gaussian noise, blurring, and resizing), and higher for another group of attacks (i.e., histogram equalization, contrast adjustment, gamma correction, scaling, rotation, and cropping).