Research On Fingerprint Watermarking Algorithm For Copyright Protection And Identity Authentication

With the rapid development of computer technology and multimedia technology, the demand for improving the information security is further increasing. The research on copyright protection of digital multimedia works and identity authentication of electronic commerce, which not only can ensure the authenticity, integrity and validity of the multimedia information,but also can provide important technical support for the protection of the security of personal property and the stability of financial transactions. On the basis of previous research results, the digital watermarking technology and fingerprint identification technology for copyright protection and identity authentication are deeply studied in this paper. The main tasks are as follows:Firstly, a double watermarking algorithm based on support vector machine(SVM) in complex contourlet domain and Krawtchouk moment is proposed. First the lower order Krawtchouk moment invariants of the blue component, which extracted from the RGB host image, are calculated for construction of robust watermarking. Then the green component image is decomposed by complex contourlet transform. Meanwhile, the watermarking image is embedded and extracted according to the SVM prediction model. Experimental results show that, compared with the digital watermarking algorithm based on wavelet domain or contourlet domain, proposed algorithm has better robustness to common image processing.Secondly, an image watermarking algorithm robust to geometric attacks in non-subsampled contourlet transform(NSCT) domain is discussed, in which scale-invariant feature transform(SIFT)key-points are used to correct geometric attacks, non-negative matrix factorization(NMF) is combined with singular value decomposition(SVD). NSCT of RGB host image is performed which the red component and blue component of the low-frequency part are extracted, respectively. Then the singular values of the non-negative matrix are obtained by NMF decomposition and SVD decomposition in the blue low-frequency component for the embedding. Meanwhile, the SIFT key-points of the red low-frequency component are used to correct geometric attacks for host image,the watermarking synchronization information is recovered and the watermarking is extracted.Experimental results show that, on the premise of invisibility, the proposed algorithm can effectively resist the geometric attacks.And then, a fingerprint image enhancement algorithm in non-subsampled contourlet transform domain based on multi-scale Retinex(MSR) is implemented. A fingerprint image is decomposedthrough NSCT. A low-frequency component and several high-frequency components in different directions are produced. Then the low-frequency component is enhanced by the MSR algorithm with hybrid intensity transfer function. While the coefficients of high-frequency components are adjusted by nonlinear gain function. Experimental results show that, compared with four enhancement methods such as histogram double equalization algorithm, multi-scale Retinex algorithm, non-subsampled contourlet transform algorithm and stationary wavelet transform and Retinex algorithm, this algorithm can improve the information entropy and contrast more efficiently, and enhances the whole visual effects. What's more, it provide support for the follow-up of fingerprint image segmentation and feature extraction.Subsequently, a fingerprint matching algorithm based on chaotic bee colony(CBC) optimization and variable boundary box is given. The geometric transformation parameters of the fingerprint matching feature are estimated by the chaotic bee colony optimization algorithm, and they are used for initial matching. The suitable matching points are selected by the threshold, which can be used for second screening through the variable boundary box. Experience results show that, compared with the fingerprint matching algorithm based on local characterist and fingerprint matching algorithm based on genetic algorithm, the proposed method has higher matching accuracy and less computation time.Finally, a design of identity authentication system including a fingerprint images fragile watermarking algorithm based on QR decomposition and a fingerprint matching algorithm is studied.The host image is decomposed through NSCT and the low-frequency parts are divided into blocks,from which the upper triangular matrix is extracted through QR decomposition. After encryption by Logistic and Arnold, the watermark is embedded into the upper triangular matrix. In the part of the authentication system, the identification of the user identity is realized by matching the fingerprint from the watermarking part and the fingerprint, which from the database or collected in field.Experimental results show that this algorithm is very good for common image processing and geometric attacks. With the fingerprint matching, the system has a higher level of security, which can satisfy the user identity authentication in the financial domain.