A Research On Mesh Simplification And Surface Subdivision

In computer graphics and virtual reality, multi-resolution analysis, three-dimensional animations, medical image reconstruction, surface modeling in engineering, CAD and other areas, three-dimensional models for large objects often require a substantial number of triangular meshes to in detail describe the characteristics of the surface for the objects. Huge amount of data brings great difficulties to computer processing and storage, network data transmission and real-time display. It is necessary to simplify the three-dimensional models. Meanwhile, the requests in computer vision for reality, real-time and interaction are increasingly enhancing, and the objects have an obvious tendency toward diversity, particularity and complex topology. How to visually account the objects has become a urgent need in the application of computer animation and human modeling software applications. Therefore it is also necessary to refine the simplified model in order to improve accuracy and smoothness of the models surface to enhance visual effects.This paper researched the theories, applications and valid resolutions of mesh simplification and surface subdivision based on new technologies and achievements of computer graphics and image processing, CAD, graphic display. Based on Garland's quadric error metrics(QEM), an improved mesh simplification algorithm is proposed. Discrete curvature and triangular area are used in the collapse cost calculation to guide the order of edge collapse. Both curvature near vertices and surface geometric features are considered in the simplification computation. To simplify the original model, the collapsing order of the large curvature region is postponed to ensure that the mesh collapses mainly to the direction of smaller curvature, making the key feature of the model preserved; the areas of triangular grid of is also introduced to the quadratic error measure matrix so as to retain the grid with larger area if they are in the same curvature. Meanwhile, edges of the mesh are not collapsed in order to reserve the feature of the object's boundary. It is demonstrated that the proposed algorithm has the same efficiency compared to the original algorithm, and meshes are distributed evenly in the simplification model, also important features of object are preserved. On this basis, the initial control mesh is subdivided to get a vivid subdivision surfaces with high precision, smoothness and good visual effects.