ABSTRACT
In order to quantitatively describe the relationship between 2D image structures and their corresponding real world structures, methods for extracting 3D information have to be investigated. According to (Marr 1982) we see 3D vision as a 3D object reconstruction task to describe 3D shape in a co-ordinate system independent of the viewer. The shape of a 3D object is represented by a 3D model. Two main classes of 3D models are identifi ed: volumetric models represent the “inside” of a 3D object explicitly, while surface models use only object surfaces. Unlike 3D models, depth maps or range images describe relative distances from the viewer of surfaces detected in the scene.
