Research Of Digital Image Zero-watermarking Algorithms Resisting Geometric Attacks

With the rapid development of computers and networks, digital media bring us lotsof convenient, but it also brings security, and other copyright issues. Being an effectivemethod for copyright protection of digital information, digital watermarking hasreceived extensive attention.Based on the analysis of the existing digital watermarking stratagy, two zerowatermark algorithms are proposed to improve the robustness of image watermarkingagainst geometric distortion:(1) A Zero-watermark algorithm based on Hu invariant moment in DCT domain.According to the copyright of the digital image, this paper puts forward azero-watermarking algorithm based on Hu invariant moments in DCT domain. First,get the8×8DCT block of the image in Lab space in brightness channel, then constructthe invariant moment, using the weights of linear array to create watermarkinformation, finally, register watermark in the third party, at the same time, build timestamp and signature. Using the specific threshold, compared with the watermark in IPRdatabase, to determine who is the copyright holder of the image. The experimentalresults show that the algorithm improves the robustness and security of the airspacezero watermark algorithm, overcame the implementation complexity and other issueswhen use traditional watermark to protect digital images, it have a higher detectionaccuracy, and it is suitable for copyright protection of digital images.(2) A blind grayscale zero-watermark scheme based on LPM and DFT. In order toresist the geometric attacks, a zero-watermark scheme which do not need watermarksynchronization was proposed in this paper. Combined with the characteristics of Logpolar mapping (LPM) and Discrete Fourier transform (DFT), the rotation, scale and translation invariants(RST) are calculated to realize the ability of images to resistgeometric attacks. In order to ensure the security and invisibility of the gray-levelwatermark, singular value decomposition(SVD) is used to preprocess the gray-levelwatermark image, then zero-watermark is embedded into the image. It is ablind-extracting algorithm because it does not require original image and originalwatermark in extraction. The experimental results show that with this algorithm, thecomplexity of geometric correction and invariant domain algorithm has been reduced,the robustness and accuracy of detection is better when compare with existing method.