Digital Watermarking Technology Based On Fuzzy Support Vector Machines And Fuzzy Clustering Analysis

Digital watermarking technology has been widely used to solve the problem of copyright protection and copyright disputes. Similar as the money watermark, it is a technology to make the special invisible mark be embedded into digital media (images, sound and video, etc.) with digital embedding method, which can be used to prove the ownership of the original author to their works, and can be the evidence of identification and prosecution of illegal infringement.Robustness and imperceptibility are the two most basic features of digital watermarking technology, but at now some algorithms can hardly satisfy the balance between these two features. So some domestic scholars begin to study the application of the mathematical analysis and machine learning method in digital watermarking technology. Based on the context, two new watermarking algorithms based on fuzzy support vector machines and fuzzy clustering analysis are proposed in this article and the main innovations in this paper are as following:1.The same membership is assigned to each type of data in the training of the support vector machines and it is hard to distinguish the importance of the different data samples; To improve the classification rate of the important class (because more information can be embedded into the important data to improve the robustness in the premise of ensuring better imperceptibility) ,bigger fuzzy membership is assigned to the important data and then watermark embedding positions are generated adaptively.2. Watermarking algorithm in DCT domain based on fuzzy clustering analysis is proposed. In order to improve the robustness and imperceptibility, the embedding positions and embedding strength are generated based on the human vision model and the idea of fuzzy clustering; then in DCT domain, the watermark is embedded into the intermediate frequency which has the better performance of hiding information. The method solves the problem of bad robustness in spatial domain.