Monte Carlo simulation for global illumination

Abstract

The aim of global illumination is to produce a realistic image that is indistinguishable from a real image of that scene in the least amount of time. This is achieved by modelling the physical laws that govern light propagation. Global illumination effects have been used in films and computer games to create more immersive environments. Unfortunately, the time taken to render a physically correct image is far higher than with more conventional three dimensional computer graphics. This not because the processes are very different but because more evaluations of certain operations such as visibility determ ination are required. The high computational costs can be addressed by either improving the efficiency of the algorithms or by parallelising them. Many of the current global illumination algorithms are limited in the types of scenes they can render and the surface models th a t can be used. Usually an importance sampling function has to be designed for each surface model. Without this function the algorithm will not perform well. In addition, as scenes get more complex, the performance of some algorithms begins to falter.