Towards the single-shot 3D movie camera

We will present new considerations and experiments aiming for the 3D movie camera – a 3D sensor which is able to acquire free-viewpoint-movies of non-rigid, moving objects, e.g. walking persons. There are already several approaches for a 3D movie camera by the computer vision-community. These approaches largely exploit temporal and/or lateral context information as well as other modalities like color to generate 3D data. Our approach is more puristic and is derived from a fundamental information-theoretical question: Is it possible to acquire a dense 3D model in one single camera frame without relying on codification methods? Our solution is based on multi-line triangulation (F. Willomitzer et al., Appl. Opt. 54 (2015) 408-417): the object surface is measured with the best possible resolution, and, due to the single-shot ability, the system acquires extensively dense 3D models with the frame rate of the used camera. This enables the possibility of high resolution 3D movies, even for high speed applications such as real-time deformation measurements in crash tests. We will show demonstration experiments.