Research On Digital Image Watermarking Algorithm Against Geometric Attack

As an effective complement means of the traditional encryption, digital watermarking is considered as an important means of digital copyright protection in the digital age and become a research hotspot in the field of information security. Digital watermarking technology refers to the hidden tag embedded into the digital multimedia data can be extracted only through a dedicated detector. As a means of data authentication and copyright protection, digital watermarking will certainly suffer various forms of attacks in practical applications. Accordingly, the measure method of the digital watermarking technology is advanced and reliable depends on whether they can be against various attacks. However, most of watermarking algorithm can not resist even small geometric attacks. Therefore, the research of digital watermarking against geometric attacks has important theoretical and practical significance.Based on previous academic research, my study of image watermarking against geometric attacks on the watermark synchronization problem has been done in this thesis. The study focuses on static image and aims at improving the robustness and invisibility of watermark. The main achievements of this thesis consist of following aspects.1. According to the masking properties of human visual system, an adaptive robust watermarking algorithm was proposed to improve the watermarking invisibility and robustness. When watermark was embedded, a DCT domain masking of original image was generated according to the properties of human visual system, and then adjusted a pair of the 8×8 DCT block intermediate frequency coefficient's relationship to embed the watermark adaptively according to its DCT domain masking value. When watermarking was extracted, we used the Zernike moment method for geometric correction to resist geometric attacks. The experimental results showed that the algorithm further improved the watermark robustness on the basis of ensuring watermark invisibility.2. Another robust watermarking algorithm was proposed, which is based on statistical properties of image histogram. In the embedded side, calculate the original image histogram firstly. Then, modify the redefined brightness range according to the value of four continuous bins samples. Finally, map the changed histogram into the original image in order to get the watermarking image. In the extracted side, also calculate the detection image histogram firstly, and then selected the same brightness interval histogram as embedded. Finally, extract the watermark by judging the relationship of four bin samples values. Experimental results show that the algorithm has both good transparency and robustness, and also can resist RST attacks, the large-portion of cropping attacks and some combination attacks.