Abstract
This paper addresses the issue of reconstructing a 3-D point from the images of entire environment projected onto a spherical surface. Two problems are mainly dealt with, they are, omnidirectional Hough transformation and intersection computation of a point to be reconstructed. A Hough transformation is very useful in extracting lines from pixel images including noise data. For the purpose of realizing omnidirectional Hough transformation, two types of models are proposed, a plane model of pixel lattice and a spherical model of voxel lattice and discuss the characteristics of these types. The processing is done in integer arithmetic. The result of an analysis shows that the two-phased spherical model gives sufficient precision. A 3-D point is computed in integer arithmetic as the intersection of two 3-D lines which are represented in Plücker coefficients. This approach is unified and robust whether the lines give a strict intersection point or not, or even a same line.
