Transform Domain-Based Digital Watermarking Algorithms Robust To Geometric Attacks

Nowadays, with the rapid development of Internet and multimedia technologies, all kinds of multimedia work (digital image, video and music, etc.) can be found on Internet. Undoubtedly, it brings conveniences to our working and daily lives, but comes along with the problem of intellectual property protection. Some dishonest person might revise and distort multimedia works on the Internet without authorized permission from the original authors. So, how to safeguard information and intellectual properties on Internet now is an urgent subject.Under this circumstance, the concept of digital watermarking, which could effectively protect the copyright of digital multimedia, came into shape in 1993. The basic idea is to embed watermarking into original multimedia works. When facing piratical cases, we can extract and compare watermarks of involved works.For more than a decade, lots of watermarking algorithms have been presented, but none of them are perfect. Most algorithms become instable under geometrical attacks, particularly under local random bending attack (RBA). Topics such as document-watermarking and multiple watermarks are also facing their bottlenecks. In my thesis, still images are taken as main research object. By utilizing the theory of image processing and character of watermarking, a feature vector robust to geometrical distortion is acquired, which can be used to embed and extract watermarking. These feature vector-based watermarking algorithms merge the character of image with encryption, and are resistant to geometric attacks.My algorithms are based on transformation domains (DCT, DWT, DFT, DWT-DCT and so on). In the end of my thesis, a set of innovative and practical watermarking scheme and performance-measuring norm is proposed.The innovations of this thesis are as follows:①Proposing a DCT-based watermarking algorithm robust to geometric and common attacks.Considering that most of the watermarking algorithms are not robust to geometric distortion, a novel algorithm is presented. It based on whole DCT. An image feature vector is acquired in DCT domain, and watermarking can be embedded and extracted by using it. The algorithm links up watermarking with image's visual character perfectly. Simulation result shows this algorithm is robust to geometric and common attacks. And it functions more effectively than the others.②Presenting a DWT-DCT based robust watermarking algorithm. This algorithm grounded on DWT and DCT. The approaching subband image is attained by using the DWT, and then the whole DCT is used for the approaching subband image. A feature vector is found from coefficients of transform domain. Results prove that the algorithm is robust to geometric attacks.③Creating a DFT-based algorithm robust to local nonlinear geometric attacks. This algorithm found on phase character of DFT. The phase character is utilized to embed and detect watermarks. This algorithm integrates phase feature with encrypting technology. Results show that this method is more robust to local nonlinear attacks than the commerce watermarking software ----- DIGIMARC, adopting in Adobe Photoshop. Comparing with other algorithms based on Fourier-Mellin, this algorithm is speedy .④Devising a DWT-based watermarking algorithm for binary images. This method is based upon DWT. In the approaching subband image, we embed and extract watermarks. Comparing with the common document watermarking methods, which usually changing the line space or word space to embed watermarks, this algorithm can hide more information, and remain unconverted to geometric distortions. The algorithm fuses the distillates of the first generation and the second generation watermarking techniques.⑤Advancing an Internet based robust watermarking system model, and bringing forward performance-measuring norm of the watermarking system.The presented model is based on transform domain, in which we find a feature vector. This system can effectively solve not only the conflicts of invisibility and robustness, but also the conflicts of invisibility and multiple watermarks.In order to measure watermarking system's performance, a quantitative index, by means of the concepts of Robustness, Speediness and Exactness, are proposed. This kind of index is formerly used only in the field of automatic control theory.In short, some creative and practical algorithms are detailed in this thesis. Results show that these proposed algorithms are robust to geometric and local nonlinear. They can effectively solve the problem of document watermarking, multiple watermarks, and they can also protect the copyrights of those un-watermarked multimedia works, which had already been uploaded onto the Internet.