Accelerating physically-based light transport algorithms
John C. Hart, Jared Hoberock · 2008
This dissertation investigates novel methods of accelerating the central computations of physically-based light transport algorithms. We present novel mappings of the crucial primitive operations of ray tracing and shading to data parallel architectures and introduce intelligent importance sampling strategies which adapt themselves to image content in an unbiased manner. We demonstrate that these techniques offer substantial improvement over the prior art and offer sufficient generality to be deployed in many path tracing-based light transport algorithms.