全局遮挡图  

Global Occlusion Map

作　　者：任重[1] 华炜[1] 鲍虎军[1] 彭群生[1] 

机构地区：[1]浙江大学CAD&CG国家重点实验室,杭州310027

出　　处：《计算机学报》2005年第6期972-978,共7页Chinese Journal of Computers

基　　金：国家自然科学基金(60203014);国家创新群体科学基金(60021201);国家"九七三"重点基础研究发展规划项目基金(2002CB312102)资助

摘　　要：基于一种有效的全局可见性的紧凑数据表示形式———全局遮挡图,文章给出了一种新的针对视点活动区域的遮挡剔除算法.对一个视点活动区域,全局遮挡图表示了一组位于空间各个方向上的可见性临界面.这些可见性临界面提供了一个不可见性判据———凡是位于该临界面后的物体必是不可见的.全局遮挡图的优点在于:(1)大小与场景复杂度无关,仅取决于空间方向的离散精度;(2)可用深度图像的方式压缩存储;(3)判断遮挡性时计算量小.同时,文章提出一种自适应构建全局遮挡图的方法,该方法综合利用了图像空间和景物空间技术的优势,适用于普遍的三维场景模型.该文在一个城市场景中对算法效率进行了测试,结果显示遮挡剔除算法可有效地提高绘制效率.Based on a effective and compact global visibility representation: Global Occlusion Map (GOM), this paper presents a novel viewcell-based occlusion culling algorithm. With respect to a viewcell, GOM in nature is a collection of directional visibility barriers, behind which any objects are definitely invisible. This nature facilitates the occlusion culling. The prominent characteristics of GOM include: (1) The storage size is independent on scene complexity, instead it merely depends on the resolution of discretization of direction space; (2) it is compressible by u-sing depth image compression technique. (3) The cost for runtime occlusion determination is rather low. Moreover, this paper proposes an adaptive GOM construction algorithm, which exploits the advantage of both image space related techniques and object space related ones, and is proved to be capable of handling generic 3D scene model. The authors test the occlusion culling algorithm in a urban scene, and show that rendering can be accelerated by a fact of 40 in regions where there are plenty of occlusions.

关 键 词：可见性剔除 遮挡剔除 全局可见性 绘制系统 可能可见面 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]