Towards Optimal Algorithms for Image Synthesis

Talk by Christian Lessig.


Numerical algorithms for image generation have myriad applications, be it for entertainment, such as for Hollywood movies and computer games, for augmented and virtual reality, or for austere ends, such as in computer aided design and architecture. In this talk, I will present our recent work allowing, for the first time, the generation of images with a minimal number of samples while facilitating bounds on the error in the computed image. I will also discuss why the development of such an optimal technique requires a thorough understanding of the physics and mathematics of image formation and its manifestation in an image. Some other applications of my research combining insights from mathematics, physics and computer science for questions in computer graphics and beyond will also be presented.


Christian Lessig is currently a postdoctoral fellow in the computer graphics group at TU Berlin. Prior, he was with the Department of Computing and Mathematical Sciences at the California Institute of Technology, which followed his Ph.D. research at the University of Toronto. Christian works on the development of verifiably good numerical simulation techniques for image generation and other physical phenomena. This requires him to combines concepts and tools from physics, mathematics, and computer science---interdisciplinary work he sincerely enjoys.