Interactive Global Illumination in Complex and Highly Occluded Environments

Abstract

Global illumination algorithms have traditionally been very time consuming and were only suitable for off-line computations. Recent research in realtime ray tracing has improved global illumination performance to allow for illumination updates at interactive rates. However, both the traditional off-line and the new interactive systems show significant limitations when dealing with realistically complex scenes containing millions of surfaces, thousands of light sources, and a high degree of occlusion. In this paper, we present an importance sampling technique that has specifically been designed for such environments. Our method maintains a rough estimate of the importance of each light source with respect to the current view using a crude path tracing step. This estimate is then used to focus computations to the most important light sources. In addition to speeding up the computation our approach minimizes the working set of the ray tracer by only touching geometry that is relevant to the current view. This allows us to directly and efficiently render scenes such as entire buildings with many thousands of light sources at interactive rates with full global illumination.