Research Of Watermarking Technology Resist To Geometric Attacks Based On Local Feature

Digital watermarking technology, which is new aspect of information hiding domain, has been widely applied by embedding copyright information into the host data. Geometric attacks which is effective attacks with its simpleness and strong offensive, is considered as the bottleneck of using robust digital watermarking technology in commercial applications. To disign a robust and secure digital watermarking scheme standing against geometrical attacks is regarded as one of the most difficult and challenging work. Study of digital watermarking resisting geometric attacks is a significant work.Through analysis the effect of geometric attacks on the image and the robustness of watermarking algorithm, this paper focuses on the feature-based image watermarking techniques and combines image feature with the template, whose advantage makes up for the shortcomings of feature. Firstly, the template information is used to correct the non uniform scaling attacks. Then the Harris-Laplace detector is utilized to extract steady feature points from the host image, the geometric invariant regions are constructed adaptively according to the feature scale. Finally, the watermark information is embedded into the geometric invariant regions using cirque quantization. Experimental results show the proposed scheme is robust against the geometric attacks, especially strong enhance the robustness to non uniform scaling attack.A contourlet-domain watermarking algorithm against geometric attacks is presented for the robustness against common signals processing and the watermarking capacity. Firstly, Harris-Laplace operator extracts steady feature points from the host image. Therefore the geometrically invariant space constructed adaptively according to the feature scale and image normalization. Finally, Contourlet transform is performed on the normalization regions and the watermark can be embedded into the bigger coefficients of the intermediate frequency directional sub-bands by using parity quantification. Experimental results show the proposed algorithm has good transparency and is very robust to common image processing and geometric attacks.