Research On Robust Digital Image Dual Watermarking Algorithms Based On DWT-SVD

The Internet has developed into the information era of cloud. Because of the faster 4G network and the intelligent devices the life of information flow is changing, the multimedia information acquisition channel is getting wider and wider, the image as a convenient form of information carrier is scattered across every corner of the Internet, so it turns out that the copyright protection of digital image has caught unprecedented attention, image watermarking technology has always been the valued field of the researchers. The digital image watermarking technology refers to hiding information in the image, so as to identify the copyright of digital image, assure authenticity and integrity authentication of digital images.In this thesis, we introduce the research background of digital watermarking, expounds the present situation of two research hotspot of digital image watermarking at home and abroad including the transform domain algorithm and anti-geometric attack and compression attack watermark algorithm. We describe the principle and classification of watermarking technology, introduces the theories involved in this algorithm in detail, and summarizes the common watermarking attack types and performance indicators.On the basis of the above research, we proposes a robust dual watermarking algorithm based on DWT-SVD, the first watermarking algorithm utilize the connection between Gauss Pyramid Distribution in the extraction process of SIFT feature and the multi-resolution characteristics of wavelet transform, and embed the watermarking information in 1, 2 or 3 corresponding to the Gauss Pyramid Octave. The exact position is determined by the approximate eigenvalue of SIFT. This algorithm ensures the confidentiality and high resistance to geometric attacks because of using SIFT values. Compared with the traditional watermarking algorithm based on DWT-SVD, this algorithm greatly improves the watermarking secrecy, perfect the ability of resistance to geometric attacks, and eliminates the false alarm. The second layer of watermarking algorithm belongs to zero watermarking. It first divides the low frequency sub bands of the DWT into blocks, and utilize the comparison result of the largest singular value of adjacent block to make a binary matrix, and then imply XOR operation to the binary matrix and the logo image to generate zero watermark registration. This algorithm solves the traditional watermark’s problem of the contradiction between invisibility and robustness of watermarking. The dual watermarking algorithm presents strong confidentiality and robustness. The first layer algorithm works as confirmation while the second as recognition, and the two algorithm are both independent and correlated, forming the strong robust dual watermarking.In the aspect of application, after summarizing the application of compressed sensing theory in digital watermarking technology, this paper use the compressed sensing as a watermark attack methods.and a number of experiments are carried out. The compressed sensing’s attack to watermarking refers to the destruction of the watermark information in the process of compression perception of the carrier image. The experiment includes the compression attack to the carrier image’s spatial domain, DCT domain and DWT domain. It shows that the dual watermarking algorithm in this paper can resist the compressed sensing attack with the compression ratio low to 0.3.