Research On Geometrically Robust Digital Image Watermarking Algorithm

The number of the watermarking algorithms proposed by researchers in digital watermarking field has been increasing year by year since recent more than ten years and digital watermarking technique made rapid progress. Watermarking is closely linked with more and more fields such as copyright protection, content authentication, note mark, usage control and it is most commonly used in copyright protection field. Robust watermarking plays a decisive role in copyright protection with its robustness to the common attacks such as JPEG compression, cropping and added noise, as well as the geometric attacks such as translation, rotation and scaling. The research on digital watermarking puts emphasis on the hard problem in the design of geometrically robust watermarking algorithms.Invisibility, embedding capacity and robustness are three main norms of the watermarking scheme and every two of these norms are mutual restrictive, so the design of the watermarking scheme simultaneously reaching the three norms is a challenging problem. In addition, on the basis of the achievement of these norms, it is of important significance to research on how to reduce the time complexity and space complexity of the watermarking scheme. Study on the watermarking focuses on finding some effective ways to solve these problems, thus design the watermarking systems with better compressive performance.Using discrete orthogonality and favorable reconstruction property of Krawtchouk moments, the robustness to the cropping of lower Krawtchouk moments including low frequency components are combined to modify some selected moments according to the size of the watermarking, image. The modified moments are reconstructed to embed the watermarking and obtain the watermarked image. The Krawtchouk invariant moments based invariant feature matrix is used to produce the authentication information for the purpose of the robustness of the watermarking. Experimental results show good invisibility, embedding capacity and robustness of the algorithm. With its stable robustness, the algorithm is suitable for the environments with high requirement of robustness.With certain level wavelet decomposition according to the size of the watermarking image, after modifying some low frequency coefficients robust to the interference, the modified coefficients are applied to obtain the watermarked image by wavelet reconstruction. The authentication information is produced by the invariant feature matrix based on the improved invariant centroid of the low frequency sub-image for its higher immunity to the attacks. Experimental results show that compared with the Krawtchouk moments based algorithm, the algorithm has higher operation speed without affecting the invisibility, embedding capacity and robustness of the watermarking and it is suitable for real-time environments.Two robust binary image watermarking algorithms are proposed by respectively applying the methods of extracting the Krawtchouk moments and the invariant centroid to binary images. Experimental results show the effectiveness of the two algorithms. In addition, using the good property that the probability density of the target pixels in the circle region of the binary image is invariant to geometric attacks such as translation, rotation and scaling, authentication information can be produced by the probability density. Experimental results show that the new algorithm reaches the main evaluation norms of the excellent watermarking algorithms. Compared with the invariant centroid based algorithm, the operation speed of the new algorithm is further increased significantly.According to the research related to the geometrically robust watermarking, progress in both theoretical and practical ways are made and formed fundamental achievement for further research on watermarking and the design of lage scale practical watermarking systems.