A Hybrid Monte Carlo Method for Accurate and Efficient Subsurface Scattering
Loading...
Date
2005
Journal Title
Journal ISSN
Volume Title
Publisher
The Eurographics Association
Abstract
Subsurface scattering is a fundamental aspect of surface appearance responsible for the characteristic look of many materials. Monte Carlo path tracing techniques can be employed with high accuracy to simulate the scattering of light inside a translucent object, albeit at the cost of long computation times. In a seminal work, Jensen et al. [JMLH01] presented a more efficient technique to simulate subsurface scattering that, while producing accurate results for translucent, optically-thick, materials, exhibits artifacts for semi-transparent, optically-thin, ones, especially in regions of high-curvature. This paper presents a hybrid Monte Carlo technique capable of simulating a wide range of materials exhibiting subsurface scattering, from translucent to semi-transparent ones, with an accuracy comparable to Monte Carlo techniques but at a much lower computational cost. Our approach utilizes a Monte Carlo path tracing approach for the first several scattering events, in order to estimate the directional-diffuse component of subsurface scattering, and switches to a dipole diffusion approximation only when the path penetrates deeply enough into the surface. By combining the accuracy of Monte Carlo integration with the efficiency of the dipole diffusion approximation, our hybrid method produces results as accurate as full Monte Carlo simulations at a speed comparable to the Jensen et al. approximation, thus extending its usefulness to a much wider range of materials.
Description
@inproceedings{:10.2312/EGWR/EGSR05/283-290,
booktitle = {Eurographics Symposium on Rendering (2005)},
editor = {Kavita Bala and Philip Dutre},
title = {{A Hybrid Monte Carlo Method for Accurate and Efficient Subsurface Scattering}},
author = {Li, Hongsong and Pellacini, Fabio and Torrance, Kenneth E.},
year = {2005},
publisher = {The Eurographics Association},
ISSN = {1727-3463},
ISBN = {3-905673-23-1},
DOI = {/10.2312/EGWR/EGSR05/283-290}
}