Optimized ray tracing for real time use without hardware accelerators

Abstract

Many real time applications such as games or animation engines wish to simulate realistic light transport to present the user with streams of high fidelity images. For decades rasterization was the rendering technique of choice because it is fast and produces high quality images. However, in recent years graphics hardware has begun including specialized cores to accelerate ray tracing, an alternative rendering solution to rasterization and a type of light transport simulation. Ray tracing is generally the more desirable choice due to its ability to present images with higher fidelity than rasterization, and due to advances in graphics hardware it can now be simulated in real time. However, with the price increases of graphics processing hardware outpacing their performance increases, and with many other devices still not including this specialized hardware due to power constraints, we present an optimized model which uses ray tracing to produce a higher quality output than that of a raster engine while at the same time not requiring the use of specialized hardware accelerators to achieve real time frame rates. In this thesis we present two experiments we performed in pursuit of this goal, one which simulated global illumination but could not maintain reasonable levels of accuracy, and another which simulated lighting, shadows, and reflections and produced images which closely resembled our reference model while achieving consistent real time frame rates.