Toral Automorphisms Mapping Frequency Domain Digital Watermarking Algorithm

With the increasing convenience of the distribution of digital products and the rapid development of the Internet, the protection of digital products is becoming more and more pressing. A watermarking system will be an effective and valuable tool to solve it.Watermarks can be applied to digital data such as still images, audio, video, etc. This dissertation considers digital grayscale image exclusively. In general, a watermark system must satisfy the two basic requirements: Imperceptibility and Robustness. In this thesis, we proposed a frequency-based watermarking algorithm using Toral Automorphisms, in which a kind of Toral Automorphism is employed to mix a recognizable binary image (BW), which contains the copyright information, into a noise-like signal. And then the mixed BW is secretly embedded into the frequency domain of the host image based on a quantization function that acts on the transform coefficients of the host image without visually degrading its quality. The reconstruction of the embedded BW doesn't resort to the original image, and the reconstructed BW is visually recognizable. Experiments show that the proposed system can successfully survive the JPEG image processing with a quality of higher than 6%, which is much lower than 10%, which most watermarking systems in the literatures can survive higher than. The embedded BW also successfully survives cropping, filtering, noise and multi-watermarking etc.The new ideas the thesis proposed are: First, frequency-based watermarking idea using Toral Automorphisms is really a novel one, although, in the literature, there are some waterniarking schemes using Toral Autpmorphisms, all of which are based on space; Second, some improvements of the quantization function contribute to the algorithm's performances to some degree.