Neuronale Priors für streuende Medien: Modellierung, Rendering und Rekonstruktion

mediaTUM – the media and publications repository of the Technical University Munich (Technical University Munich) 2026 Thesis Open access

Read full text
Authors: Ludwic Leonard
Cited by: 0

Abstract

This thesis advances forward and inverse rendering of participating media using neural priors and differentiable rendering. Contributions include: (1) a generative model for efficient subsurface scattering simulation; (2) a variance-reduced framework for reconstructing heterogeneous volumes; and (3) a diffusion-based method for single-view volumetric reconstruction with learned priors and physically-consistent rendering.

Fields of Study & Subjects

Computer Graphics and Visualization Techniques 3D Shape Modeling and Analysis Advanced Vision and Imaging

Publication details

Journal / source
mediaTUM – the media and publications repository of the Technical University Munich (Technical University Munich)
Publication date
26 January 2026
Language
EN
Document type
Thesis