Simplifying Mesh With Complex Attributes

With the development of 3D scanning techniques, 3D mesh models are recently becoming a new type of media after images, sound and video. It's widely used in entertainment and manufacturing and other industrial application. In order to approach better visual effect, models often have material properties such as colors, textures, and surface normal. But the traditional algorithm doesn't do well for these models. So it's significant to present a method to account for a wide range of surface attributes.This paper describes the relevant simplification algorithm, and proposes the approach for appearance-preserving simplification. The main job in this paper is as follows:1. First we produce approximations based on iterative edge contraction and quadric error metrics. But models might have a distinct piece of texture for each face, and this would create a discontinuity at every edge. And the quadric error metrics algorithm doesn't works well for this kind of model. So we present a method for handle this problem.2. The thesis uses normal map to refine the coarse models. Firstly we sample the coarse models to get a series of sample points, then find the corresponding points on the original models, finally use corresponding points'normal substitute for the samples'when we render the coarse models. By this method, coarse models can appear more geometric details, just like color and texture. To realize texture mapping, we partition an arbitrary surface into many disk-like charts, and then use harmonic map to parameterize mesh into squares, at last sort them with a rectangular packing of the greedy algorithm in the texture space, which improve and optimize the utilization of the texture.3. The thesis uses selective refinement to refine the region where the users interested. By this way, we can restore the region's attributes, and during the process of region refinement we also can modify charts which generated by partitioning mesh models.We firstly introduce the conception of the surface simplification and then present an overview on the algorithm for simplification, especially the classic simplification methods. Second we propose the method for surface segmentation and parametric. Third we refine in the region in order to restore the details that people interested and then update the charts dynamically. Forth, we propose how to save the algorithm of model to simplify, increase more surface details by normal mapping, and pack the square charts to form a texture atlas by greedy algorithm. At last we talk about the future work and the problem we confront now.