Research On Robust Digital Image Watermarking Algorithms

With the development of the digital techniques and Internet, it becomes so easierfor the copy and broadcast of the digital works. The regrettable one is that moreand more digital works are used illegally. In order to safeguard the copyrightowner's legitimate rights and interests, we must take the effective measure toreduce or stop the using illegally of digital works and to reduce the economiclosses caused by this. Digital watermarking, as a kind of effective copyrightprotection means, it stresses the advantage is that the digital watermark isimperceptible and does not influence the usability of the multimedia data;Secondly, the digital watermark is closely related to digital works protected, verydifficult prerequisite destroying the works protected shifts down except that thewatermark. Hence it has already become one of the important technologies forcopyright protection. The application of digital watermarking technology is verywide, such as broadcast monitoring for digital TV, copy control for DVD, contentverification for digital camera, and anti-counterfeit of different certificates et al.The researches on digital watermarking involve multi-disciplinary theory andtechnology of field, for instance information theory, code theory, communicationprinciple, cryptography, etc. The studies on digital watermarking technology haveimportant academic meanings, and extremely important economic meaning andsocial effect no matter for angles of theory and application.The basic principle of digital watermarking is to utilize redundancy existing inthe multimedia data, and embed watermark information that can be distinguishedinto the digital works to be protected. The watermark can be extracted whileneeding in order to finish copyright protection, etc. This dissertationsystematically overviews the current status and development trend of digitalwatermarking technology, especially robust digital image watermarking. Thisdissertation mainly investigates the robust digital image watermarking algorithms.On one hand, several modified watermarking algorithms are presented afteranalyzing the virtues and shortcomings the existing watermark algorithms;On theother hand, several novel watermarking algorithms are proposed by combining thenew technology and theory in the signal processing field. Main innovativecontributions are as following:1. The general issues of robust image watermarking are discussed, and oursettlement to enhance the robustness of image watermarking is presented. Wesystematically summarized several factors influencing the digital watermarkingfor the first time. On the basis of further investigating, analyzing many kinds ofrobust digital watermark algorithms, we propose the possible solution to improvethe robustness of the digital image watermarking.2. To the deficiency of the existing blind watermark algorithms, a novel watermarkingalgorithm based on Independent Component Analysis (ICA) is proposed. TheSingular Value Decomposition (SVD) is employed during the embedding processof the watermarking. Instead of random sequences, some readable signature isused as watermark. By means of ICA, the watermark is successfully extracted. Inthe proposed algorithm the capacity of the embedded watermark is enhanced, andat the same time the transparency can be guaranteed. Our method can extract thewatermark without taking attacks into consideration. Experiment results havedemonstrated that the proposed approach is robust against the common signalprocessing, such as JPEG compression, additive noise. In particular, it is alsorobust against geometric attacks, such as rotation, cropping, scaling etc.3. To the shortcoming of the existing watermarking robust to geometricaldistortions , especially the rotation distortion, a robust image watermarkingtechnique is presented. The watermarked image is obtained by modifying themaximum singular value in each image block. The robustness of the proposedalgorithm is achieved from two aspects: the stability of the maximum singularvalues and preprocessing before watermark extraction. Zernike moments are usedto estimate the rotation angle and the translation and scaling distortions arecorrected by geometric moment methods. Experimental results show that thisalgorithm makes a trade-off among the imperceptibility, robustness and capacity.4. As there is no existing watermarking scheme using for copyright protectionand copyright tracing, a blind image watermarking algorithm based on HT(Hadamard Transform) is proposed. Different watermarks are embedded into theimage at the phase of image production and release to the aim of copyright tracing.Based on the comparison of the characteristics HT, DCT and DFT, HT domain isselected for watermarking. After partition the original image into sub-blocks, HTis implemented. The AC coefficients with Zigzag order are the positions towatermarking. The image watermark is firstly Arnold transformed to enhance thesecurity and robustness. Experimental results showed that the proposedwatermarking algorithm is imperceptible and robust to common signal processingoperations and geometrical distortions.5. The Contourlet transform is a novel geometrical image-based transform,recently introduced by Do and Vetterli. The Contourlet transform is adopted byvirtual of its advantages s over the wavelet transform in order to obtain a flexiblemultiresolution, local, and directional image expansion using contour segments.The watermark is inserted through multiplicative embedding. We present amaximum likelihood watermark detection method based on modeling theContourlet coefficients with Generalized Gaussian Distribution (GGD) havingzero mean. Under the Neyman-Pearson criterion, the decision rule is optimized byminimizing the probability of missing the watermark for a given false detectionrate. Experimental results demonstrate the invisibility, and robustness to signalprocessing. What made people more joyful is, this algorithm can resist theprinting and scanning effectively.