Research On Intelligent Computing Techniques Of Geometric Distortion In Digital Image Watermarking

The applications of multimedia content in the Internet environment have grown rapidly in two decades,it makes copyright protection an important issue to be solved urgently.As one significant means of copyright protection,robust watermarking technology has become a research hotspot in the field of information security.During the study of robust watermarking technology,how to resist geometric attacks effectively is still a difficult problem,especially for rotation and scaling.Therefore,it is of vital research value and practical significance to deliberate on robust watermarking algorithms against geometric attacks.At the beginning of this dissertation,i study the problem that geometric transformations destroy the synchronization of digital watermark embedding and extracting processes.Based on the Polar Harmonic Transforms(PHTs),evolutionary optimization and Support Vector Machines(SVM),a robust and intelligent watermarking scheme against geometric attacks is designed.The main contents of research and specific work are as follows:(1)A digital image watermarking algorithm based on PHTs moment is proposed aiming at the problem of the watermarking synchronization distortion caused by geometric attacks such as rotation and scaling.The watermark embedding process includes three main steps:first,orthogonal moment features are calculated using the invariance of PHTs moments;then the stable moment invariants that are insensitive to geometric attacks are selected according to the order and repeatability of the PHTs moments;finally,the watermark is embedded by quantizing the selected stable moments,and the original image and the compensated image are reconstructed by combining the unmodulated and modulated PHTs moments respectively,eventually the watermarked image is obtained.The experimental results show that the proposed watermarking method in resisting geometric attacks(rotation and scaling)and general signal processing attacks(gaussian filtering,median filtering,average filtering,gaussian noise,salt and pepper noise,speckle noise,JPEG compression,etc.)has a superior performance,compared with the existing corresponding watermarking methods,(2)Combining two intelligent computation methods:Differential Evolution(DE)algorithm and SVM,an intelligent image watermarking scheme against geometric attacks is proposed.Based on the watermark algorithm mentioned in(1),the DE algorithm is utilized to search for optimal watermarking quantization step,so the balance of relationship between imperceptibility and robustness can be kept greatly.The SVM is used to train detection model of the watermark based on its optimal classification ability,so our method can realize intelligent extraction of watermark bits information.The experimental results show that the proposed watermarking scheme can significantly achieve high detection rate.(3)In order to explore the research results further,the print-scan attack is considered as an important test for the proposed watermarking method.After analyzing how the print-scan process affects digital image quality and watermark extraction,the robustness of the PHTs moment towards print-scan attack is verified through abundant experimental results.The test of the print-scan attack is performed on the watermarking scheme proposed in(1),the experimental results show that the proposed method is effective for copyright protection of image in print-scan environment.