The Research Of Color Image Watermarking Algorithm Against Geometric Attacks

Digital watermarking technology is a comprehensive subject of communication, pattern recognition and information security. Researchers from different fields study in different angles, forming the situation of a hundred flowers blossom and a hundred schools of thought contend. The urgent need of digital copyright protection and information security makes watermark robustness as a hot issue of digital watermarking technology.The important standard of measuring watermark's robustness is whether the watermarking can resist various attacks. At present, digital watermarking system resist attacks can not achieve people's expect requirements. In general, digital watermarking algorithm takes on robust for noise, compression and filtering attacks, but against geometric attacks is not mature enough. The characteristics of geometric attacks is when the image attacked by geometric slightly, the watermarking is not synchronism with the carrier, then making the detector failed to detect it. Therefore, how to synchronize the watermark with the carrier image is the key technology of against geometric attacks. For this problem, we do the follow works:1. Discussed digital watermark theory, and analyzed related theoretical knowledge of watermark embedding space, watermark embedding domain and watermark security, made bedding for the follow-up study of watermarking algorithms against geometric attacks2 Using singular values' stability realized the watermark synchronize with the carrier image, given a color image watermarking algorithm against geometric attacks based on DWT-SVD. Combined with the wavelet theory, CIEL*a*b* color space and singular value decomposition theory, more copies of the same watermark are embedded in carrier different singular value of wavelet sub-band coefficient, to improve the ability of watermarking against geometric attacks. Then, we do the simulation of the proposed watermarking system in Matlab7.0 experimental platform, realized embedding watermark into the carrier image and extracting watermark from the watermarked image. The experimental results show that the two algorithms proposed in this thesis are robust on normal image processing attacks and especially on geometric attacks.3. Using the characteristics of the image feature points can still maintain after geometric attacks, realized the watermarking synchronize with the carrier image, given a color image watermarking algorithm against geometric attacks based on local image features (SIFT feature points). Extract sift feature points in the color image and use this feature points construct Delaunay triangulation net; Then, embed watermarking in the wavelet domain of one triangle. fill the triangle into a rule square triangle, then embed watermarking in its wavelet domain; then, we do the simulation of the proposed watermarking system in Matlab7.0 experimental platform, realized embedding watermark into the carrier image and extracting watermark from the watermarked image, and verified watermark's robustness on common geometric attacks.In the experiments, watermark is a 2-dimensional binary image, which carries large amount of information, costs less and easy realize. Study results show that the algorithms designed in this paper have better robustness and invisibility, and can well against geometric attacks.