Volumetric object reconstruction using generalized voxel coloring

Abstract

This paper presents a volumetric approach for three-dimensional (3D) object reconstruction. Building 3D models of objects from 2D images is still a very difficult task. However, the built 3D models can have different applications in many areas, such as in industrial inspection, virtual reality, and medicine, among others. In our work, an uncalibrated turntable image sequence of the object to be reconstructed is acquired. Then, 3D reconstruction is performed using Generalized Voxel Coloring (GVC), a volumetric method based on photo-consistency criteria. Two experimental objects, with different topologies, were reconstructed. The robustness of the GVC method was verified by varying some parameters, like the number of images used and the influence of the objects silhouettes considered.