Research And Implementation Of Text Copyright Protection System

Nowadays, copyright works such as microblog and novel, which are mainlystored in digital forms and disseminated in paper forms by printing and scanning,often generate disputes. Digital watermarking technology is an effective way tosolve copyright disputes. However, the research of text copyright in the digitalwatermark field is still in beginning stage because of the poor redundancy of textdocument. The existing text watermarking algorithms have the problems of weakrobustness or low embedding capacity. This dissertation researched digitalwatermarking algorithm for text copyright protection in both text files and textimages, which presented a new text similarity computation method and anoptimization watermarking algorithm based on odd-even quantization.For text files, two similarity computation methods based on feature itemsand feature frequency were presented to solve the less robustness problem on thebasis of zero-watermark algorithm. Firstly, the operation of preprocessing, featureextracting and dimensionality reduction were completed for text files and the itemfrequency of feature items were obtained by the statistics. Then, the similarity offeature items and feature frequency were calculated according to the presentedmethod to determine whether the copyright was tort by comparing with apredefined threshold. Thirdly, a text copyright protection system was designedwhich could process the formatted and unformatted texts. Using the system, textfiles copyright protection could be achieved for Chinese text, English text and the mixed text. Meanwhile, the system also had many functions, such as batchoperations of multiple files, the report of files with copyright infringement, andestablishing the platform for international graphic user interface.To solve the problem of low embedding watermark capacity for text images,the dissertation proposed the odd-even quantization watermarking algorithmbased on printing and scanning invariables. First of all, the preprocessingoperations for text images, such as binaryzation and image patch, were done.Secondly, an optimization quantization function was presented on the basis ofprinting and scanning invariables. Then, the flipping pixels strategy wasemployed for each image patch based on human visual masking characteristics.The similarity calculation method and the odd-even quantizationwatermarking algorithm presented in this dissertation were experimented. Fortext files, experimental results indicate that the proposed similarity computationmethod is more robust and stabile compared with other algorithms. Even if thescale of text contents that suffered from modification attack to50%, the similarityis still more than70%. For text images, because the optimized algorithm has fullyconsidered the masking characteristics of human visual system, experimentalresults show that the optimized text image watermarking algorithm has goodimperceptibility and enlarges the watermarking capacity on the premise of strongrobustness.