Research On Digital Watermarking Of 3D Mesh Models

With the rapid development of Internet, virtual reality, and cooperative design, more and more 3D digital products are entering the World Wide Web. The owners of the digital products face more and more seriously tortuous activities, for example unauthorized occupancy, duplication, modification and so on. So the watermarking techniques of 3D mesh models are becoming solutions of this problem. As an effective solution for copyright protection and authentication, the watermarking technique of 3D mesh models is becoming a new hotspot after the watermarking technique of texts, images, audios, videos. This dissertation researches and discusses the robust watermarking techniques for copyright authentication of 3D mesh models.First, the dissertation make a survey of the related knowledge about the digital watermarking including the background, basic idea, applications etc. The principle, characteristics, classifications and performance measures are presented in detail. Based on the typical watermarking schemes at home and abroad, we research the frequency domain watermarking schemes. The main works we made is as follows:A new improved frequency domain watermarking scheme has been proposed. Before embedded watermarking, the 3D mesh model is affinely transformed into a new 3D model, which is robustness to translation, rotation and scaling. Then the distance of each vertex in the mesh is calculated to the origin of the mesh, and the real number sequence of distance is transformed to a signal in frequency domain using one-dimensional discrete cosine transform (DCT) . Watermark are enbedded into the signal in frequency domain, and the embedded watermark to be imperceptible. This scheme is robust against common mesh attacks such as mesh simplification, affine attack, mesh cropping, random noise added and combined attacks.Based on the researches about feature point and partial distortion of 3D mesh model, an adaptive watermarking scheme of 3D mesh model is proposed based on feature point. It finds out the points from the high variation area, and looks them as the feature point, which can be embedded watermarks. The feature points are grouped into different section according to their modular .Then the real number sequences of the sections are transformed to a signal in frequency domain by using one-dimensional discrete cosine transform . The proposed algorithm is not only robust to affine attack, but also robust against common mesh attacks such as mesh simplification, mesh cropping, random noise added and combined attacks. At the same time, it can effectively reduce the partial distortion extent.