We describe a method to use Spherical Gaussians with free directions and arbitrary sharpness and amplitude to approximate the precomputed local light field for any point on a surface in a scene. This allows for a high-quality reconstruction of these light fields in a manner that can be used to render the surfaces with precomputed global illumination in real-time with very low cost both in memory and performance. We also extend this concept to represent the illumination-weighted environment visibility, allowing for high-quality reflections of the distant environment with both surface-material properties and visibility taken into account. We treat obtaining the Spherical Gaussians as an optimization problem for which we train a Convolutional Neural Network to produce appropriate values for each of the Spherical Gaussians' parameters. We define this CNN in such a way that the produced parameters can be interpolated between adjacent local light fields while keeping the illumination in the intermediate points coherent.
The paper can be found at https://doi.org/10.1111/cgf.13918.
You can download the author version of the paper from here.
@article{SphericalGaussianLightfields:currius:2020,
author = {Currius, R. R. and Dolonius, D. and Assarsson, U. and Sintorn, E.},
title = {Spherical Gaussian Light-field Textures for Fast Precomputed Global Illumination},
journal = {Computer Graphics Forum},
volume = {39},
number = {2},
pages = {133-146},
doi = {10.1111/cgf.13918},
year = {2020}
}