Hidden Information Encoded Digital Image Information-based Security Research

With the rapid and extensive growth of computers and the advent of Internet, our everyday life and work are becoming digitalized and network-connected. Huge amounts of multimedia content are available over the network, and increasingly commercial transactions are performed through it. The classical copyright protecting ways and security techniques have been challenged because they couldn't protect the digital information thoroughly. In order to satisfy the security needs, protect the copyright of owners, and better fit to the typical characters of digital multimedia, scientists and researchers have devoted to various technical solutions. Digital watermarking studied deeply and detailed in this paper is one of these techniques.Digital watermarking embeds some mathematical information into the multimedia data (such as image, audio, video signals etc.) in order to protect the copyright. Generally it demands robustness and invisible.First, this thesis introduced digital watermark in detail, give a common framework of watermarking system, and demonstrate the general methods of watermark production, embedding and detection. Then, some important research results of watermarking are presented and the attacks towards watermarking system are discussed.The emphasis of this thesis is the author's research and innovation in the field of watermarking. Based on the previous work and current research situations, this thesis has completed the following work:At first, wavelet is introduced and a new adaptive watermarking scheme is proposed. Experimental results show that this scheme enhance the imperceptibility of watermark and is robust to most attack Especially to EZW (Embedded Image Coding Using Zerotreesof Wavelet) compression.Another new watermarking method proposed in this paper, applies SVD (singular value decomposition) to the watermark image and the Walsh transformation to sub-blocks of the host image. In this method, many lesser singular values arediscarded and the largers which bear the relatively information are embedded repeatedly. When extracting the watermark, the data can be checked by the statistic of the repeated data. And the outputs are the average of the believable parts. The experiments result shows that the proposed method is robust to many common attack, especially the cropping attack..The third digital image watermarking method bases on FastICA (Fast Independent Component Analysis). In this method, the host image is transformed into discrete wavelet domain. A binary image is embedded to the approaching sub-image. When extracting the watermark, a FastICA method is employed. The experiments show that the algorithm has good transparency and robustness.Digital watermarking is a new technology, which refers to communication and information technology, image and speech processing, signal detection and evaluation, mathematics, human audio and visual systems, computer network and application etc. Although in recent years great progress has been made in the aspects of theory and applications, there is no complete theoretical framework, especially no universal specification for judgment. There is still some problems need to been solved. We can say that digital watermarking is a field full of vagueness and challenges. This paper makes an outlook for digital watermarking in the end.The digital watermark mentioned here is only for digital image, those for audio and video are not concerned in this thesis.