We present a new approach for accelerated global illumination computation in scenes with glossy surfaces. Our algorithm combines sparse illumination computation used in the radiance caching algorithm with BRDF importance sampling. To make this approach feasible, we extend the idea of lazy illumination evaluation, used in the caching approaches, from the spatial to the directional domain. Using importance sampling allows us to apply caching not only on low-gloss but also on shiny materials with high-frequency BRDFs, for which the radiance caching algorithm breaks down.

Thanks to its ability to improve the realism of computer-generated imagery, the use of global illumination has recently become widespread among digital lighting artists. It remains unclear, though, what impact it has on the lighting design workflows, especially for novice users. In this paper we present a user study which investigates the use of global illumination, large area lights, and non-physical fill lights in lighting design tasks, where 26 novice subjects design lighting with these tools. The collected data suggest that global illumination is not significantly harder to control for novice users that direct illumination, and when given the possibility, most users opt to use it in their designs. The use of global illumination together with large area lights leads to simpler lighting setups with fewer non-physical fill lights. Interestingly, global illumination does not supersede fill lights: users still include them into their globally illuminated lighting setups. We believe that our results will find use in the development of lighting design tools for non-expert users.

Local Principal Component Analysis (LPCA) is one of the popular techniques for dimensionality reduction and data compression of large data sets encountered in computer graphics. The LPCA algorithm is a variant of kmeans clustering where the repetitive classification of high dimensional data points to their nearest cluster leads to long execution times. The focus of this paper is on improving the efficiency and accuracy of LPCA. We propose a novel SortCluster LPCA algorithm that significantly reduces the cost of the point-cluster classification stage, achieving a speed-up of up to 20. To improve the approximation accuracy, we investigate different initialization schemes for LPCA and find that the k-means++ algorithm [AV07] yields best results, however at a high computation cost. We show that similar ideas that lead to the efficiency of our SortCluster LPCA algorithm can be used to accelerate k-means++. The resulting initialization algorithm is faster than purely random seeding while producing substantially more accurate data approximation.