3D Volumetric Reconstruction and Characterization of Objects from Uncalibrated Images

Abstract

Three-dimensional (3D) object reconstruction using only bi-dimensional (2D) images has been a major research topic in Computer Vision. However, it is still a complex problem to solve, when automation, speed and precision are required. In the work presented in this paper, we developed a computational platform with the main purpose of building 3D geometric models from uncalibrated images of objects. Simplicity and automation were our major guidelines to choose volumetric reconstruction methods, such as Generalized Voxel Coloring. This method uses photo-consistency measures to build an accurate 3D geometric model, without imposing any kind of restrictions on the relative motion between the camera used and the object to be reconstructed. Our final goal is to use our computational platform in building and characterize human external anatomical shapes using a single off-the-shelf camera.