Research On Digital Watermarking And Its Application

In recent years the research on digital watermarking is always a hot topic in the field of computer applications. Usually the watermark is required to be strong enough to undergo all kinds of signal processing operations. Its two key problems are the robustness and the security, which are used to improve the viability of the watermark in different conditions. And the difference between them is whether digital watermarking is permitted to detect or not. Our research work is done just about the two issues.The task of the research on the watermark security is to prevent watermark from being removed or modified maliciously when the watermark detection method is public. In this thesis, we propose a layered asymmetric watermarking method, which combines some features in symmetric watermarking and cryptography. On the basis of it, we have also developed a layered asymmetric watermarking system, in which the two strong requirements of an asymmetric watermarking system, i.e. robustness and non-invertibility, are put into different layers of structures. And the ultimate watermarking result is given in the upper layer. By using an asymmetric encryption the attacker is able to detect the watermark in the upper layer but unable to change it. In the lower layer, we use an improved SWE algorithm from Wong et al.'s to modulate the watermark in the direction of the host vector to get a better robustness. And the ECC coding technique is also applied to further improve the lower layer's reliability. The experiment results show that our digital watermarking scheme is good in robustness, reliability and invisibility. And its security is also strengthened by the asymmetry of the upper layer.The aim to the research on the robustness is to enhance the watermark's ability to resist all kinds of removal attacks. In this thesis, we propose a novel watermark-attack algorithm, which is based on fractal code principle. The digital watermark can be effectively removed by using our algorithm and it is better than other image coding techniques (such as JPEG compression). According to the experiment results, we found that the watermark has only a little effect on Hausdorff metric when using IFS code. Our algorithm tries to look for some similar image blocks to replace those image blocks with the particular watermarks subjected to contract mapping and Hausdorff metric according to the fractal IFS theory. It also tries to weaken the effectof watermark on the image attacked so that the watermark is very difficult to be detected. We have finished our experiments on all different kinds of images. The results show that the watermark correlation output degrades effectively in both attacks to wavelet domain robust watermarking and DCT domain robust watermarking. Downgrade of image quality is analyzed and compensated. The experiments further indicate that a better attack effect can be gained by combining the attack proposed in the thesis with other signal processing operations. The work may provide some new metrics in watermark robustness evaluation, and also urge research works on watermarking attack to progress forward.