Computations on a spherical view space for efficient planning of viewpoints in 3-D object modeling

Viewpoint planning plays an important role in automatic 3D model generation. Previously (H, Zha et al., 1997), we proposed a viewpoint planning method to determine the next-best-viewpoint (NBV) for incremental model construction. Based on a current partial model, this algorithm provides quantitative evaluations on the suitability of viewpoints as the NBV. Since the evaluation is performed for all potential viewpoints, the NBV planning is very time-consuming. We present a novel method of discretizing a spherical view space by a look-up array which will highly facilitate the NBV evaluations. Two main issues are addressed: 1) a uniform tessellation of the spherical space and its mapping onto the 2D array; 2) incremental updating computations far evaluating viewpoints as the NBV. The efficiency of the method is verified by algorithmic analyses and experiments using a real modeling system.