A Digital Watermark For Vector Graphics Based On Visual Features

With the convenience the Internet has brought during the process of the exchange of digital information, the digital information can also be illegally obtained, copied and modified very easily. Therefore, the information security and copyright protection of the digital products become a crucial problem. Digital watermarking provides the digital products with effective protection. Different from the information encryption technology which just restricts the acquisition, the protection watermarking brought the digital products with can resist both the illegal transmission and the modification during the application. Though the main research areas of the digital watermarking are mainly image, audio and video with developed techniques, digital watermarking in graphics, especially 2D vector graphics is newly emerging. Contrast with the 3D mesh which has high commercial value, the protection of the 2D vector graphics has gained little attention. However, as the important data format in GIS and the cartoon with high commercial value, it is also meaningful to protect the 2D vector graphics.In this thesis, the basic principle and the domestic and oversea research situation of digital watermarking technology are introduced briefly. According to study of the fully developed image watermarking techniques, a systematic understanding of the basic framework of digital watermarking technology is got. And the watermarking technology to 2D vector graphics is focused. Then establish the definition of base contour with the feature points of shapes, and implement the separation of the base contour from the original contour based on the theory of shape representation. With all the techniques mentioned above, a blind watermarking algorithm applied in 2D vector graphics is proposed. The watermark is embedded in the amplitude of the normalized difference between the Fourier Descriptors of base contour and the original contour. Then the watermarked contour is produced via contrary operation. Because of the application of the feature points, the detection of watermarking of the algorithm can be done without the original contour, that is, the algorithm is blind. With the concept of the low frequency of information is used in shape as base contour in the algorithm, the embedding of the watermark modified little of the most important parts which are sensitive by human visual, so the watermark is unperceivable with high robustness, which accomplishes the objective of copyright protection. At last, the experiment results of the algorithm are given to test the robustness of the algorithm.The experiment results show that, the watermarking algorithm proposed in this thesis is perceptual transparent and robust against a variety of attacks such as changing the starting vertex, translation, rotation, uniform scaling, noise and simplification at a small extent.