An efficient statistical data representation for real-time rendering of metallic effect car paints

Abstract

Realistic virtual reality applications require highly-detailed geometry as well es convincing surface representations. In many applications, especially in the automotive industry, the realistic rendering of metallic effect paints is necessary. Due to their complex appearance, this is a demanding problem. Previous methods either use a computationally heavy-weight and often hand-tuned simulation approach or a data-driven approach. The former are thus not well-suited for real-time applications. The latter have the advantage of lower computational complexity and virtually no manual hand-tuning, but the disadvantage of requiring large amounts of the graphics card’s memory, making them problematic for larger scenes with numerous materials as required in VR applications. In this paper, we describe an efficient representation for metallic car paints, based on computing the statistical properties of measured real-world samples. Our approach is suited for real-time rendering, poses only moderate requirements on the computing power, uses a low amount of memory and displays high-quality results, as shown in our evaluation section. As an additional advantage, our representation allows the generation of BTFs of arbitrary resolution.