Volumetric Video Capture

Since 2010, we developed and investigated multi-view capture systems with the aim to capture and reconstruct full-body size person. We started with a wall-display like configuration (figure left). In 2016 we extended our camera setups towards 200 degrees capture angle around a person (figure right).

For full body 360 degree performance capture, an even more complex setup was required. In October 2017, we started to conceptualize, develop the first 360 degree capture stage. This new studio concept for the recording of complete volumetric 360° video objects was put into operation in prototype form in the institute's own studio and, based on this, a continuous workflow for volumetric video was implemented. The recording system is equipped with 32 high-quality cameras, which are arranged as 16 stereo pairs and each have a very high local resolution of 20 MPixels with a temporal resolution of 30 frames per second. These stereo pairs are mounted at three different heights to record the person from below, from the middle and from above. The necessary increase in the number of cameras and the use of high-resolution sensors also brought new challenges for data processing. When recording with this system, a raw data volume of 1.6 TByte per minute is created, which then has to be converted into time-consistent, dynamic 3D models with photorealistic textures in the subsequent 3D video processing.