【作者】 詹海生；

【导师】 周利华；

【作者基本信息】 西安电子科技大学 ， 计算机应用技术， 2006， 博士

【摘要】 二维折线集、三角网格和点是几何模型表示中的基本元素,分别被广泛用于二维矢量地图和三维物体模型的表示。如何快速地传输和操作由它们表示的几何模型是实时图形处理的主要任务。本文以几何压缩理论与方法为基础,分别分析了矢量地图和三维几何模型中的拓扑数据和几何数据的特点,首先,提出了基于折线简化和单纯复形的矢量地图多分辨率表示和渐进式传输方法,该方法在保持视觉一致性的条件下,将较小分辨率地图的数据量降低为最大分辨率地图的数据量的1/10,提高了矢量地图渐进传输的效率。其次,提出了基于扇形带的三角网格拓扑关系表示方法,该方法在不考虑顶点分裂与融合的情况下,可以保证连接关系的最低压缩率为log 2 (274)比特/顶点。第三,提出了基于广义三角形带的三角网格数据压缩和多分辨率表示方法,该方法可以充分发挥图形处理器的功能,统一了单分辨率和多分辨率三角网格的表示方法。最后,提出了基于KD-树的点模型各向异性量化方法。对KD-树广度优先的遍历可以构造出多分辨率点模型。该方法可以将点与多边形相结合混合表示几何模型,克服了点模型在表示大平面时效率低的缺点。以上方法有效地提高了图形在存储、传输以及绘制等方面的效率,增强了图形处理的实时性。更多还原

【Abstract】 The two-dimension polyline set, triangle mesh and point sampled model are primary methods of geometric model representation and are widely employed to represent the vector map and three-dimension geometric model corresponsively. How to transmit and operate these geometric models is the important task of real-time graphics. Basing on the principle and methodologies of geometric compression, we analyzed the properties of the topological and geometric data contained in vector map and proposed new strategies for saving, transmitting and rendering these data as well as triangle mesh and point sampled model. The first strategies presented in this dissertation is the multiresolution representation and progressive transmission of vector map based on polyline simplification and Simplicial Multi-Complex by which the volume of coarser level data can generally be reduced to one tenth of that of the finest level while preserving the visual consistency between them and the progressive transmission of vector map data is speeded up. The second is the fan strip based topological data representation for simple triangle mesh by which the connectivity between vertice can be encoded less than log 2 (274) bit/vertex when the split vertex and merge vertex is ignored. The third is the generalized triangle strip based data compression and multiresolution representation for simple triangle mesh. The single and multi resolution representations have same codec strategies in this method which is appropriate to the architecture of modern graphic process unit in personal computer. At last, KD-tree based anisotropic quantization was developed to reduce the volume of geometric data, i.e. coordinates of point sample, for a 2-manifold point sampled model. The breadth-first traversal of KD-tree was used to construct the multiresolution presentation. In this method, the hybrid point and polygon rendering surmounts the obstacle of large flat surface representation by point sample set. All these strategies increase the efficiency of realtime graphics processing.更多还原