Abstract
Digital watermarking, a branch in the field of information hiding, has drawn lots of attention of researchers and has become an important technique in various application areas. Watermarking has been widely used for content protection, copyright management, content authentication and tamper detection etc. Various kinds of watermark techniques have been proposed in the past decade, and a key problem of this issue is how to balance the tradeoff between watermark robustness and the fidelity of images.Based on the way to embed the watermark sequences, existing techniques can be categorized into two fields: one is to embed the watermark in spatial domain directly, and the other is to hide it in the frequency or transform domains. Today, digital watermarking techniques with better confidentiality are those developed in transform domains. Hence, in this thesis, we adopt typical transformed-based watermarking methods, such as based on DCT- and DWT-based schemes, to design and implement our watermarking algorithm. As for the robustness issue, we have to recognize that no matter how the robustness is measured, the pros. and cons. of a transform-based watermark scheme depends on the characteristics of the transform it used. In this thesis, we propose a joint watermarking technique and derive the way to “integrate” two independent watermarking approaches, such as DCT- and DWT-based ones. With this idea we can ensure the hiding data would be found in both transform domains without interference. Meanwhile, we will show that even the host data suffered from malicious attacks, our watermarks can still be detected in at least one of the transform domains; that is to say, they could be complement to each other. Finally, several experiment results, including JPEG compression, cropping, scaling, Gaussian noise, rotation, translation, composite attacks, stirmark and so on, will show that the proposed method has good performances on confidentiality and robustness.The breakthroughs of our work are the watermark derivation and the watermark detection performance strategy. By calculating the true positive rate and false positive rate we show the ROC curve utilizing Spread Spectrum and SVM method, so that they will help us to diagnose the performance of our watermarking scheme.