Date of Award


Document Type

Thesis (Undergraduate)


Department of Computer Science

First Advisor

Wojciech Jarosz


In computer graphics (especially in offline rendering), the current state of the art rendering techniques utilize Monte Carlo integration to simulate light and calculate the value of each pixel in order to generate a realistic-looking image. Monte Carlo integration is a highly efficient method to estimate an integral that scales extremely well to a high number of dimensions, making it well suited for graphics, because generating images creates a high-dimensional integrand. The efficiency of these Monte Carlo integrations depends on the sampling techniques used, and using a more efficient sampling technique can make a Monte Carlo simulation converge to the right answer quicker than using more naive sampling techniques. In this thesis, we present an efficient sampling method that demonstrates much higher performance than many other sampling techniques. This novel sampling method, based on orthogonal arrays, offers guaranteed stratification in arbitrary projections, leading to better theoretical performance with integrands that have cross-correlated variance compared to sampling methods that do not offer these same guarantees.


Originally posted in the Dartmouth College Computer Science Technical Report Series, number TR2019-872.