Berlin – Decai Chen, former research scientist in the Immersive Media & Communication group of the Vision and Imaging Technologies (VIT) departement at Fraunhofer HHI successfully defended his PhD dissertation entitled “Deep Learning-based Scene Reconstruction from Multi-view Videos” at the Chair of Visual Computing at Humboldt University in Berlin in June 2026.
In his dissertation, Decai Chen addresses key challenges in reconstructing dynamic 3D and 4D scenes from multi-camera video recordings. His research bridges classical multi-view stereo techniques and modern deep learning approaches to advance the state of the art in volumetric video capture and reconstruction.
The dissertation presents several significant scientific contributions:
- Developed a deep learning-based multi-view stereo pipeline for highly accurate human reconstruction from volumetric video captures.
- Introduced novel neural surface reconstruction techniques that improve geometric fidelity and rendering quality.
- Proposed a template-free framework for reconstructing complex dynamic scenes and human-object interactions.
- Created an efficient 3D Gaussian Surfel approach enabling fast and temporally consistent 4D scene reconstruction.
- Advanced the state of the art in volumetric video by improving reconstruction accuracy, temporal consistency, and computational efficiency.
During his time at Fraunhofer HHI, Decai Chen made important contributions to research in Visual Computing and volumetric video technologies. His successful PhD defence marks a significant milestone in his scientific career and highlights the strong connection between academic excellence and applied research.
The dissertation was supervised by Prof. Dr. Peter Eisert, Chair of Visual Computing at Humboldt University and Head of the Vision and Imaging Technologies Department at Fraunhofer HHI.
Fraunhofer HHI warmly congratulates Decai Chen on this outstanding achievement and wishes him continued success in his future scientific and professional endeavours.