Research On Algorithms For Digital Watermarking And Fingerprint Recognition In Copyright Protection And Authentication

With the extensive application of multimedia technology, more and more digital works need tobe encrypted, authenticated and copyright protected. How to prevent information leakage of digitalworks and how to confirm the copyright of owners when the information leakage occurs are veryimportant. Therefore, copyright protection and authentication of digital works have become animminently practical problem. On the basis of analysis and summary of research results at homeand abroad, this thesis studies digital watermarking and fingerprint recognition algorithms incopyright protection and authentication. The main tasks are as follows:Firstly, a watermarking algorithm based on kernel fuzzy C-means clustering, singular valuedecomposition and human visual model in the complex wavelet domain is studied. First, texturesand backgrounds of carrier images related to human visual system are processed by kernel fuzzyclustering. Then, singular value decomposition is conducted on the low-frequency coefficients ofcomplex wavelet decomposition of carrier image. Finally, the watermark is embedded into thesingular values. Experimental results show that the algorithm has good robustness and invisibility;it has better performance than the watermarking algorithms in spatial domain or wavelet domain.Secondly, the watermarking algorithm against geometric attacks based on scale invariantfeature transform (SIFT), pseudo-Zernike moments and complex contourlet transform is proposed.Low-frequency approximation components of complex contourlet transformation are chosen as thesample population for the SIFT algorithm. The local interest feature point range of the image isconstructed by key points. Pseudo-Zernike moments of the range are calculated and normalized.Watermark is embedded by jitter quantitative moment amplitude method for each local interestfeature point range. The experimental results show that the algorithm has strong robustness forgeometric attacks.Then, a fast iterative algorithm of improved two-dimensional minimum Tsallis cross-entropythresholding based on gray level-gradient histogram is implemented for image segmentation. Byusing the gray level-gradient histogram and its regional division method, the fast iterativealgorithmic formulas of minimum Tsallis cross-entropy thresholding method based on the twodimentional histogram regional divisions are derived.The image thresholding segmentation isachieved. Experimental results show that the algorithm has better performance and less computingtime than the original cross-entropy algorithm and Tsallis cross-entropy algorithm based on gray level-average gray level histogram.Next, a fingerprint feature matching algorithm based on point pattern and support vectorclassification is discussed. First, the fingerprint images are matched by fingerprint feature of pointpattern. Then, support vector machine is adopted for the classification training of sample vector,and the detail minutiae are matched again. At last, the two matching results are fused by classifier.The experimental results show that the algorithm has better performance than the fingerprintfeature matching algorithm based on traditional point pattern or point pattern of local feature.Finally, a fingerprint images fragile watermarking algorithm based on Logistic chaotic mapand a design of identity authentication system are discussed. In the fragile watermarking algorithmpart, low-frequency components are first extracted from the carrier image using complex contourlettransform, and then the singular values of low-frequency components are calculated. At last, theauthentication watermark is generated by Logistic chaotic map, which is embedded into thesingular values. In the identity authentication system part, the fingerprint information of legal usersis embedded into ID card as authentication watermark. When it comes to personal identification,fingerprint information of ID card is extracted and matched with the collected fingerprintinformation to complete the personal identification. Watermarking algorithm shows goodvulnerability and imperceptibility, and improves the security of automatic fingerprint identificationsystem (AFIS) since the fingerprint information is not from the database, which avoids to betampered by users.