A Watermarking Scheme Based On DCT And SVD

As the computer communication technology is rapidly developing, the transmission of digital multimedia information becomes more convenient. The internet has brought a lot of chances to the commerce, scientific research and entertainment via some new kinds of service, such as electronic publishing, electronic advertisement, data warehouse, digital library, internet video and audio, e-commerce. On the other hand, it also brings us an urgent and important problem that how to protect the copyright of multimedia data. The techniques of encryption and scrambling can ensure the security of the product to some extent. However, the authorized digital products will be illegally copied and transmitted once they are decrypted. As a new technology, the digital watermarking method has been proposed to avoid tortious. Its essential is to implement the information hiding by utilizing the image processing. The key point is embedding the watermark into the host image and simultaneously maintaining its transparency. The watermark, which bears the information of copyright and certification, can identify the author, owner, publisher and user and thus guarantee the legal copy and transmission.We first briefly introduce the concept of watermark and its background, and emphasize its importance in the application. Then we review in some detail the exact definition of watermark, its classification and characteristics. Second, the existing digital watermarking models are discussed and divided into three categories, which correspond to the three stages of the watermarking technique, i.e. embedding, extraction and testing, respectively. We detailedly elucidate the algorithms of watermarking, such as spatial domain and frequency domain transformation, and the Singular Value Decomposition (SVD). In addition, the preprocessing of the image can also play an important role in some watermarking algorithms. Therefore, we enumerate several methods of image scrambling, such as Arnold cat mapping, baker, Hilbert curve and Mobius strip transformations.In this thesis, the watermarking algorithms based on Discrete Cosine Transformation (DCT) and SVD are amply analyzed. In order to improve the previous works, we put forward a novel watermarking method, which combines the Arnold image scrambling technique and DCT-SVD algorithms. In this new algorithm, the characteristics of DCT (resistance against the compression), SVD (robustness) and preprocessing of the watermark via Arnold image scrambling (resistance against the cropping) have been gathered together and proved to be very useful in practice. Finally, we introduce various attacks of digital images and give the extracted watermark after the action of these attacks. In comparison with the algorithms in the literature, we conclude the advantages and shortcomings of our method. It turns out that this DCT-SVD method combined with Arnold cat mapping has certain robustness against the attacks including JPEG compression, Mediate filter, Gaussian noise and rotating and cropping.