3D Content Creation by Passive Optical Methods

The possibility of obtaining 3D models, i.e., mathematical descriptions of real objects or scenes has paved the way to a wide range of new and exciting applications in fields such as virtual simulation, human-computer interaction (HCI), scientific visualization, cultural heritage documentation, medicine, industrial prototyping, reverse engineering, entertainment (movies and video games), web-based commerce and marketing, just to name a few. The construction of the 3D model of a real object or scene by optical sensors, also referred to as 3D modeling pipeline, essentially consists of four steps: 1) data acquisition, 2) calibration, 3) reconstruction, and 4) model editing. Any optical sensing device used to collect data can only capture the surface front side and not what is occluded by it. Therefore, a full model must be built from a number of images covering the entire object (data acquisition). In order to perform 3D reconstruction, the camera’s parameters must be estimated by a procedure called calibration. Such information can also be obtained from the acquired images if they represent some common regions (by a procedure which is typically called self-calibration). Reconstruction is then performed and the resulting model is stored in an efficient description such as polygonal meshes, implicit surfaces, depth maps or volumetric descriptions. In practical situations, reconstruction may lead to models with some imperfections; thus, a further repairing step is recommend (model editing) [Davis et al. (2002); Levoy et al. (2000)].