Fragile Watermarking Algorithm For Text-Image Content Authentication

In our daily life, a lot of documents such as contract, notice, certificate and voucher are saved as digital images which are called text-image. The modification of digital image is not easy to be found, so, we need some technology to verify the authenticity of the image content. Digital image watermarking technology can protect the authenticity and integrity by embedding the watermark into the image. This thesis works on the research of watermark embedding and tamper detection based on fragile text-image watermarking for image authentication.First, the background and significance of fragile text-image watermarking schemes are introduced. Then, fragile binary text-image watermarking schemes are analyzed from watermark generation, watermark embedding and tamper detection. And some shortcomings of the existing algorithms are summarized.To improve the ability against collage attack, this thesis proposes a binary text-image fragile watermarking algorithm based on one-to-one block-mapping. In this scheme, the watermark of each block is hidden into the flipped pixels of the corresponding mapping block. The authenticity of each block is detected by comparing the consistency between the reconstructed watermark and the extracted one. The neighborhood tampering feature is used to improve the tamper detection performance. Experimental results show that the proposed algorithm can not only detect the modification, addition, or deletion, but also effectively resist collage attack. Based on the above studies, we noticed that there are many homogeneous regions in the binary text-image. In view of the above-mentioned facts, a new algorithm based on one-to-many block-mapping is designed. The watermark of each block is embedded into many mapping blocks. The experimental results indicate that it can effectively locate the modications on the binary text-image which have more homogeneous regions.According to the demands of electronic invoice content authentication, a color text-image fragile watermarking algorithm is proposed based on block-classification in this thesis. This scheme put the PNG color electronic invoice image as the research object. All image blocks of each component are divided into three categories, and then the watermark of each block is embedded into certain positions of the corresponding mapping block according its category. In tamper detection stage, for each component, the authenticity of each block is determined by comparing the watermark consistency and using neighborhood extension detection. And then the detection result of the test image is obtained by integrating RGB components. The simulation results show the imperceptibility of the watermark, the small file delta of watermarked image, and successful localization of different attacks.Finally, a simulation system, which is easy to operate with friendly user-interface, is designed to detect the content authentication of text-image. The system can be used to verify the validity and usefulness of the proposed algorithm through the simulation analysis of different algorithms.