基于层次聚类树的点曲面视点相关绘制  

View-Dependent Point Set Rendering Based on Hierarchical Clustering Tree

作　　者：徐爱国[1] 黄琦[1] 孙守迁[1] 

机构地区：[1]浙江大学计算机科学与技术学院,杭州310027

出　　处：《中国图象图形学报》2007年第2期356-365,共10页Journal of Image and Graphics

基　　金：国家重点基础研究发展规划"973"计划(2002CB312106);国家自然科学基金项目(60475025);浙江省科技计划(2004C33086)

摘　　要：针对点曲面的视点相关绘制问题,提出了一个新的表面基层次聚类简化算法。区别于普遍采用的空间剖分基策略,该算法的显著优势在于能够运用法向锥半角误差标准有效跟踪曲面的起伏变化,并以此为聚类简化过程提供可靠的全局误差控制。离线简化阶段,连同各种预定义的聚类约束条件,算法构造了点曲面模型的连续层次多分辨率表达。实时绘制阶段,层次可见性裁剪以及优化的树遍历提高了系统的整体性能。此外,通过引入附加的轮廓增强机制,在较大的屏幕投影误差和较高的模型简化率情况下,系统仍然能够保证较好的绘制视觉质量。This paper proposes a novel algorithm for surface-based hierarchical clustering simplification that aims to accelerate view-dependent point set rendering. The remarkable advantage of this algorithm is that it uses a normal cone semi-angle metric to trace surface curvature variation and provides the clustering simplification process with a reliable global error control, rather than commonly used spatial partition scheme. In off-line simplification phase, combined with various predefined clustering constraint conditions the algorithm constructs a continuous multi-resolution hierarchical representation for point set model. In real-time rendering phase, hierarchical visibility culling and optimized tree traversal efficiently improve system integral performance. Moreover, an additional silhouette enhancement mechanism is introduced to ensure a well rendering vision quality in spite of a larger screen projection error and higher model reduction rates.

关 键 词：模型简化 多层次细节 点基绘制 多分辨率 

分 类 号：TP391.72[自动化与计算机技术—计算机应用技术]