The ability to represent real-world scenes in a photorealistic manner with 6 degrees of freedom (DoF) promises a multitude of novel applications in entertainment, e-commerce and documentation. However, today these potentials are difficult to leverage due to the complexity of content creation. Light-fields aim to contribute to the solution of this challenge. They excel by their faithful representation of real-world scenes in a photorealistic manner.
This is explained in more detail by first discussing the role of light-fields in the ecosystem of 6DoF content production and playback. Next, we present a software solution called Realception® that processes light-field data and plays them back for virtual reality applications. We will show how the light-field data can be integrated into the Unreal Engine and how a node-based user interface entitles the user to adapt the processing to the captured content for an improved visual quality. We will conclude our talk by the experiences we made during a first test production and explain the resulting consequences on future end-to-end solutions. This addresses in particular the tradeoff between capture efforts and viewing flexibility as well as occurring data volumes and computation times.
Joachim Keinert, group head “Computational Imaging and Algorithms”, Fraunhofer Institute for Integrated Circuits
Joachim Keinert is heading the group “Computational Imaging and Algorithms” of the Fraunhofer Institute for Integrated Circuits in Erlangen, Germany. He has diplomas from both the University of Stuttgart, Germany, and Télécom ParisTech (formerly ENST) in Paris, France. In 2009, he received a Ph.D. with honors in the domain of electronic system level design for image processing applications. His main field of research is on light-field imaging and its potential for media production and for digitization of real-world scenes. Beforehand, he also worked on high-dynamic range video capture, JPEG 2000 and low complexity coding, where he actively participated in the JPEG compression standardization.