Date of Award
2022
Document Type
Thesis (Master's)
Department or Program
Computer Science
First Advisor
Wojciech Jarosz
Abstract
Traditional steady-state rendering assumes that the light transport has already reached equilibrium. In contrast, time-of-flight rendering removes this assumption and recovers the pattern of light at extremely high temporal resolutions. This novel rendering modality not only provides a way to visualize the propagation of light, but can also empower the advances in time-of-flight imaging and its corresponding applications.
Building on previous work in steady-state volumetric rendering, this thesis introduces a novel framework for deriving new Monte Carlo estimators for solving the time-of-flight rendering problem in participating media. Conceptually, our method starts with any steady-state photon primitive, like a photon plane or parallelepiped, and slices it with a temporal wavefront, producing a primitive of one dimension lower. We show how these sliced photon primitives arise by analytically integrating four dimensions in a spatio-temporal extended path space formulation. The differences in these primitives reduce to the determinant of a 4 × 4 Jacobian matrix, which results in different strengths and weaknesses. We then demonstrate the possibility of combining their strengths using multiple importance sampling. Finally, we implemented several of the new estimators enabled by our theory and compared them with existing techniques.
Our new theory holds great potential in supporting the design of more efficient estimators for volumetric time-gated rendering, and can thus benefit a wide range of corresponding imaging applications, such as providing cheap yet realistic training data for machine learning algorithms that reconstruct objects through highly scattering media, and allowing researchers to explore the trade-off between new time-of-flight imaging systems virtually.
Recommended Citation
Liu, Yang, "Temporally Sliced Photon Primitives for Volumetric Time-of-flight Rendering" (2022). Dartmouth College Master’s Theses. 52.
https://digitalcommons.dartmouth.edu/masters_theses/52
