Geometric Robust Watermarking Based On Scale-space Theroy

Along with the continuous progress of multimedia technology and the increasing prevalence of computer network, the application of digital media gains rapid development. Digital media is easier to be copied and transmitted, compared with traditional media. This advantage brings more convenience to people, but also makes the copyright protection of digital media the severest problem in present digital industry.The advent of digital watermark technology gives a hopeful way to solve the problem. The current digital watermark algorithms have good robustness against some often-used attacks such as compression, filtering, noise, etc but lack of effective resistance to geometric attacks such as rotation, scaling, cropping, translation and local random distortion etc.By deeply analyzing the influence of geometric attack on the image and the robustness of digital watermark algorithms, this paper focus on feature detection in algorithms against geometric detection and propose a feature point detecting method upon watermark system based on the theory of scale space. The detected feature points have high repeat rate under large-scale transformations and can also change with the image scale transformation. Taking the relative position of feature scale and feature point as a reference, this paper presents a watermark embedding area founded on self-adapted patterns of geometric transformation. With the transformation of feature points'relative position, the self-adapted patterns also make corresponding change, which can fulfill the purpose of resisting geometric attack.Then, according to research on nonisotropic transformation in geometric attack, this paper further spreads the application field of scale-space theory in watermark design and proposes an algorithm against geometric attack grounded on affine invariant feature point. The feature point in affine scale-space has higher repeat rate under nonisotropic geometric transformation and the ellipse window of nonuniform Gaussian kernel on characteristic scale is founded on image contents. By analyzing the geometric characteristics of ellipse window and the statistic feature of its containing image, this paper improves embedding area detecting method, as well as the robustness of embedding area detection. Based on this, a watermark embedding and detecting method is designed and realized.At last, experiments are done to validate the presented algorithms. The experimental results show that the algorithms can effectively resist various of geometric attacks, relatively improve the robustness of scaling and aspect ratio transformation and has high performance.