摘要
针对可见外壳的体构造方法需要事先知道包围盒和转成三角面表示时低曲率区域存在大量小三角面的问题,本文提出一种基于体素生长的自适应采样可见外壳构造方法.将物体三维表面视为等值面,用多目立体匹配的方法求得等值面附近的一个空间点;然后以该点为中心构造体素,将体素在等值面上生长直至遍历整个等值面;同时根据体素的曲率变化控制下一个生长体素的大小.实验结果表明本文方法在不知道物体包围盒大小的情况下能够计算出可见外壳,而自适应采样可解决过多小三角面的问题.
Focusing on the problems in visual hull construction using volumetric approaches,which need bounding box and have too many tiny triangles in low curvature region by marching cubes,a voxel growing based adaptive sampling method is presented.In this paper,visual hull surface is represented by an iso-surface,and a point in 3D space is got via multi-view stereo matching;Then,a seed voxel is constructed using the 3D point as the center which grows around the iso-surface.The growing voxels size is inverse with current seed voxel＇s curvature in a adaptive sampling manner.Experiments show that the method can construct visual hull without bounding box and the adaptive sampling can resolve problems of existing volumetric approaches which have too many tiny triangles in low curvature region.
出处
《小型微型计算机系统》
CSCD
北大核心
2010年第9期1877-1880,共4页
Journal of Chinese Computer Systems
基金
教育部-英特尔信息技术专项科研基金项目(MOE-INTEL-08-08)资助
关键词
可见外壳
三维重建
剪影
等值面
体素
自适应采样
visual hull
3D reconstruction
silhouette
iso-surface
voxel
adaptive sampling
