This thesis discusses the physical and information theoretical limits of optical 3D metrology, and, based on these principal considerations, introduces a novel single-shot 3D video camera that works close to these limits. There are serious obstacles for a perfect 3D-camera: The author explains that it is impossible to achieve a data density better than one third of the available video pixels. Available single-shot 3D cameras yet display much lower data density, because there is one more obstacle: The object surface must be encoded in a non-ambiguous way, commonly by projecting sophisticated patterns. However, encoding devours space-bandwidth and reduces the output data density. The dissertation explains how this profound dilemma of 3D metrology can be solved, exploiting just two synchronized video cameras and a static projection pattern. The introduced single-shot 3D video camera, designed for macroscopic live scenes, displays an unprecedented quality and density of the 3D point cloud. The lateral resolution and depth precision are limited only by physics. Like a hologram, each movie-frame encompasses the full 3D information about the object surface and the observation perspective can be varied while watching the 3D movie.