On reconstructing surfaces of arbitrary topology from a range image

To reconstruct a topologically consistent 3D surface model is a challenging problem. In this paper we present an algorithm that estimates from a range image, without registering and converting it into a volumetric data, the most probable topology of a surface to be reconstructed. In our algorithm we start with a closed initial mesh and preserve its "closeness" during the reconstruction process. Some experiments were carried out to validate our proposal.