Surface reconstruction and representation of 3-D scenes

Abstract This paper describes a method of reconstructing and representing the surface of objects in a 3-D scene from a set of parallel planar slices or cross sections. The 3-D scene may consist of several occluding objects. Each slice may consist of a number of distinct 2-D regions defining the 2-D structure at that particular cross-section. The method described does not require that the scene description be a completely segmented one, i.e. separate object structures need not be distinguished from each other. The method can be used where only incomplete description due to missing data are available. The generation of surface description consists of three steps. First, the contours of 2-D regions are traced and an attempt to segment the 3-D scene is made by associating contours on different slices. Second, a triangulation process is performed over pairs of associated contours to generate the local bounding surface structure. Finally, a surface hierarchy is established by coalescing adjacent triangular patches of same orientation into polygonal facets; these facets are then integrated to form the object and scene description. The surface structure thus obtained is a collection of polygonal facets which bound the outer surface of the objects.