Research On Digital Watermarking Techniques For Copyright Protection Of Three-Dimensional Mesh Models

With the rapid development of digital geometry processing, network technology and hardware equipment, three-dimensional (3D) models have been widely applied to industrial manufacturing, urban planning, architectural design, medical care, cultural heritage protection, film and television production, games and virtual reality, etc. As a digital media, 3D models can be edited, modified and stored conveniently by computer or other digital devices while they are also easy to duplicate and transmit at a low cost and without any loss of quality through a variety of storage media, computer networks or other means of data transmission. However, these advantages make it illegal to occupy, copy, edit, and disseminate 3D models that infringe the copyright of the owner becomes very easy. How to protect the copyright of a 3D model effectively and prevent it from infringements is becoming a hot issue need to be resolved. As an effective method of copyright protection, digital watermarking technology has attracted more attentions of researchers. However, most previous achievements have focused on image,audio, video and text. Because the methods used to represent a 3D model differ while the operations performed on the 3D models are varied and complex, research achievements in the area of 3D model watermarking technology still need to be further improved. The thesis focuses on the copyright protection of 3D mesh models while the goal is to improve the robustness of a digital watermarking algorithm against common attacks. Through studying the problems of embedded watermarking algorithm and zero-watermarking algorithm, the thesis has completed the following four aspects of work.(1) Digital watermarking technologies for copyright protection of 3D mesh models are summarized. The 3D digital watermarking is classified from different technical views. Then, from two aspects of embedded watermarking algorithm and zero-watermarking algorithm,the technical concepts, processing frameworks and representative achievements are stated, analyzed and generalized. We point out the advantages and deficiencies of the related work and summarize the attacks and countermeasures of a 3D model watermarking algorithm as well as the performance evaluation metrics.(2) Simplification and cropping are two common operations for a 3D model while they are also the attacks that are most difficult to resist for a watermarking algorithm. In order to improve the robustness of a embedded watermarking algorithm against the simplification, cropping and their joint attacks, a new watermarking algorithm for 3D mesh models is proposed where a global geometrical feature is used combined with model partition, non-blind detection and trend analysis by analyzing the characteristics of these attacks. According to the spherical coordinates and vertex norms of vertices,model partitions and vertex sets in each partition are established. Then, in each vertex set, based on the mean of vertex norms the number of vertices changed by algorithm is adjusted to embed a watermark bit. Non-blind detection is adopted by watermark detection procedure while the watermark is extracted through comparing the number of vertices with different changes base on trend analysis. Experiments show that the algorithm can resist more attacks than the other non-blind algorithms and improves the robustness against the simplification, cropping and their joint attacks.(3) The current zero watermarking algorithms usually use a single feature to construct a watermark that makes them only roubst against some attacks. Aiming at this problem, by analyzing the limitations of using one single feature to construct a watermark, the idea of multi-feature processing is introduced. A new zero-watermarking algorithm is proposed that uses a global geometrical feature and a local geometrical feature combined with the methods of stable vertex selection and data statistics. Stable vertices representing the basic shape of a model are identified using OSVETA.Afterwards, vertex norm partitions and SDF value partitions are built. Two skewness vectors about vertex norms and SDF values are calculated by skewness measure respectively, and a ratio vector about vertex norms is extracted. Then a watermark is constructed by using these three vectors. Experiments show that the algorithm improves the robustness of zero-watermarking against the common attacks.(4) The technical defects of the embedded watermarking and the zero watermarking make the current 3D model watermarking algorithms not robust against the single attacks having high intensity and joint attacks. Aiming at this problem,through analyzing the advantages and deficiencies of the embedded watermarking algorithm and zero-watermarking algorithm,a new watermarking algorithm for 3D mesh models by constructing vertex distribution characteristics is proposed that fuses the ideas of the two algorithms. According to the angular components of the vertex spherical coordinates, vertex partitions and vertex sets of a 3D model are established. In each vertex partition,a watermark bit is constructed according to the distribution characteristics of vertices based on the idea of zero watermarking. Subsequently, the distribution characteristics of vertices are enhanced and the watermark is embedded into the vertex sets based on the idea of embedded watermarking. In watermark detection,watermark is extracted from vertex partitions and vertex sets respectively and then the final watermark is determined by analyzing the vertex partitions. Experiments show that the algorithm improves the robustness against the single attacks having high intensity and joint attacks effectively.