根据形状类型信息由图像中的明暗变化获取物体的表面方向  

On Improving Determination of Normal to Surface in Shape from Shading(SFS) through Division of Surface into Four Types of Patches

作　　者：须明[1] 赵荣椿[1] Maria Petrou 

机构地区：[1]西北工业大学计算机学院,陕西西安710072 [2]萨里大学信息电子技术学院

出　　处：《西北工业大学学报》2005年第1期70-74,共5页Journal of Northwestern Polytechnical University

基　　金：国家自然科学基金 (6 0 14 10 2 );英国皇家学会项目 (Q775 );西北工业大学博士创新基金;南昌航院测控中心实验室基金资助

摘　　要：研究一种专门用于 SFS(由图像的明暗重构物体表面三维形状 )结合其它辅助信息的条件下 ,提取光滑物体表面方向信息的 SFS算法。物体表面上的曲面片按其方向曲率的符号组合分成 4种类型 :凸形、凹形、横向鞍形和纵向鞍形 ,以便于利用辅助信息加以实际判别。对应于这种分类方式 ,设计出相应的 4种由明暗关系计算表面方向的算法基本算子。这种新算法在算子与表面类型相匹配的情况下 ,能够取得很好的结果。In our opinion, ill-posed nature of SFS lies in the fact that a certain shading may probably correspond to several different shape meanings. We explore in this paper the possibility of eliminating ambiguity through replacing the usual assumption of dome type surface by dividing the curved surface into four types of curved surface patches: dome, cup, horizontal saddle and vertical saddle. Four surface orientation operators are designed corresponding to the four kinds of regular patches. Four types of patches and four corresponding surface orientation operators form only the core idea of our exploration. In the full paper we give details of our exploration. Three experiments show that our algorithm performs well when the proper type of curved surface patch and corresponding surface orientation operator are selected according to the details given in the full paper. It looks promising that the success of our exploration, when combined with ancillary information, may solve the ill-posed problem of SFS.

关 键 词：SFS 曲面表面方向 曲面片类型 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]