The Research Of Robust Image Watermarking Technology Against Geometric Attacks

With the development of computer network and multimedia techniques, the digital revolution brings limitless convenience for people. However, the issue of the intellectual property rights problem has become increasingly serious. Digital watermarking as an effective solution to this problem which embeds the imperceptible signal (watermarking) into the multimedia data in order to defeat against unauthorized copying and intentional tampering has attracted more and more attention in the field of information security. Recently, there are many unresolved problems in the field of digital image watermarking. Robustness is one of the problems. The watermark synchronization is very crucial scheme to design robust image watermarking. Therefore, how to develop the robust digital image watermarking schemes against geometric attacks has been the hot topic in the past years. In this paper, two novel image watermarking schemes are proposed to cope with the geometric attacks.1, a robust DFT image watermarking detection algorithm against geometric attacks based SVR. Six combined low order image moments are taken as the feature vector and the geometric transformation parameters are regarded as the training objective, then a SVR training model can be obtained. The combined moments for test image are selected as input vector, the actual output (geometric transformation parameters) is predicted by using the well trained SVR. Finally, the digital watermark is extracted from the corrected test image. Experimental results show that the algorithm is robust against common signals processing and the geometric attacks.2, a robust image watermark insertion and detection scheme which can resist to general geometric attacks based on the theory of SVR. The Pseudo-Zernike moments of the image are computed according to rotation invariability of pseudo-Zernike moments, and some low-order Pseudo-Zernike moments are selected. Finally, the digital watermark is embedded into the host image by quantizing the magnitudes of the selected Pseudo-Zernike moments. In the process of watermarking detection, the algorithm selects the low-order Krawtchouk moments as the eigenvectors of SVR. Experimental results show that the scheme is not only possess the ability of non-sensation, but also robust to common signals processing and general geometric attacks.