The Mesh Simplification Algorithm Based On 3D Segmentation

With the development of computer science and technology, more sophisticated and realistic 3D models are required in many fields, such as computer graphics, virtual reality, computer aided design, geographical information system and medical image technology. Therefore, the complexity of models is calculated by more than millions of triangles, i.e., the David model, used in digital Michelangelo project at Stanford Univeristy, can reach up to two billion triangles.Millions of triangles are required to construct such complex models—up to 2 billion triangles were used to model the sculpture of David in the Michelangelo project at the Standford University. The complicated models propose great challenges to the storage capcity, processing power, rending speed and transmitting rate of computer system. However, high resolution is not always necessary in some applications and there is a trade-off between resolution and processing time. So simplified model is used in some cases to replace complex original model. Mesh simplification, which redueces the mesh complexity, is the technology to generate such simplified models.Using proper algorithms, mesh simplification reduces the number of polygons, edges and vertices of a model while retaining its shape, appearance and other most important features. Mesh simplification plays a very important role in the storage, transmitting, processing and rendering of a model.Mesh simplification is the act of transforming a three-dimensional polygonal model into a simpler version. It reduces the number of polygons needed to represent a model while trying to retain a good approximation to the original shape and appearance. And it is very important for storage, transmission, process, and real-time rendering. In this paper, the mesh's progressive adaptive simplification is thoroughly researched based on mesh segmentation, and the main work and contributions would be addressed as follow:Considering the effect of visual sensitiveness areas, we proposed a view-dependent progressive multi-resolution mesh simplification algorithm. In this algorithm, a half edge collapse operator is introduced base on the power function, and a serie of new technologies, such as pre-determination ruler, heap sorting and local update, is adapted to improve the simplified performance.In addition, another adaptive mesh simplification approach is proposed, which is base on the meaningful mesh segmentation. Actually, people usually assume the complex object as simple basic elements combination, and tend to divided it into several parts by the region of minimize negative curvature. According to this fact, a new automatic meaningful mesh segmentation method is proposed. The method concentrates on feature contour, gotten from the minima negative curvature value, for cutting. Based on minima rule and part salience theory from the cognitive theory, my approach divides a mesh into salient parts along the concave discontinuity of the tangent plane. My novel method first extracts features to find candidate contours, then the open contours are prioritized and automatically completed to form loops around mesh's parts. The loop completion is constrained by two parallel cutting planes, which are selected from the open feature contour's Oriented Boundary Box. On the basis of the segmentation results, a various level power function is assigned to different type vertices in the before simplification algorithm, addressed in chapter two, to obtain adaptive multi-resolution.Experiments proved that the progressive mesh algorithm has the resonableness of human perception while preserving human vision perfering regions. The automatic adaptive mesh simplication algorithm based on meaningful segamentation can generate natural visual segmentation with high performance.