Watermarking Techniques For Image Content Authentication

The revolution of digital technology has increased the ease of manipulation, reproduction and distribution of digital media; meanwhile, it also leads to potential security problems. It is an important technique to verify the authenticity and integrity of digital images using content-based digital watermarking. Supported by "Watermarking and Data Hiding", the innovation project of Northwestern Polytechnic University, we address some key issues of image content protection and authentication using digital watermarking in this dissertation. The major contributions of this dissertation are as follows:1. We proposed the theoretical relationship between the length of spread spectrum watermarking and the false detection rate, which provides efficient means to calculate the watermark length of spread spectrum watermarking for image content authentication. The main idea comes from the analysis of stochastic process models of image and watermarking, as well as the statistical property of watermark correlation detection. The experimental results and the theoretical estimations meet well.2. We propose a new algorithmic scheme for content protection oriented digital watermarking (CPODW) to overcome the common problem of existing content based digital signature methods, i.e., high false positive. From the aspect of image understanding, we also present two classes of image content features for image authentication. Experiments demonstrate that the average error ratio of the CPODW decreases from 46.9% to 11.7%, compared to the traditional methods.3. We propose a novel image content based fragile watermarking authentication algorithm. The image content based watermarking is embedded into the wavelet domain, which, unlike traditional techniques, characterizes the image distortions so that malicious tampers can be distinguished from integrity-preserving operations. It improves the reliability of algorithm by conducting the tamper detection in both spatial domain and frequency domain.4. We propose a mixed blind watermarking algorithm, based on wavelet-coefficients tree, to improve the robustness of image content protection. The low and high frequency components of image are adopted with quantization and spread spectrum techniques respectively for the embedding of content watermark. The experimental results illustrate that the statistical average testing measures of our proposed method are 2.1 times more than that of spread spectrum method.5. We design a image content authentication system framework based on public key and digital...