Reconstructing a 3D model from range images using radial flow model

The reconstruction of a 3D model from range images can be conveniently split into two stages. The first stage consists basically in the extraction of geometrical information, e.g. the depth and the orientation of image points, and the second stage concentrates on the transformation of these data to a 3D model. In this paper we focus on the second stage. Our approach is based on a deformable model, which offers more flexibility to mold a 3D model according to several geometrical constraints. Moreover, instead of looking for a global functional that carries out the deformation, we opted for establishing discrete local correspondences between 3D surface point data extracted from images and 3D model points.