Real-time image-based lighting of metallic and pearlescent car paints

Abstract

The complex behavior of metallic and pearlescent paints is not accurately captured with simple (spatially varying) BRDF models. They exhibit goniochromatic effects and glints due to metallic flakes embedded in the paint. However, when viewed from a distance or when illuminated from a larger area light source, the high-frequency glints tend to vanish. Therefore, we do not consider glints in this work and focus on the remaining angular color variations of a homogeneous materials, based on the direction from which the paint is illuminated or viewed, and how to realize these effects in a real-time rendering engine. Many models have been proposed to capture these effects, but efficiently and accurately reproducing them in real-time rendering applications under image-based lighting is a challenging problem. Therefore, most of the analytical models are not suitable either because of their computational complexity or because they are not able to present these effects accurately enough. We propose a prefiltering technique to handle angular color variations with very little overhead in comparison to the widely used standard methods for real-time image-based lighting. We require that the BRDF can be factorized into a brightness BRDF which works well with real-time image-based lighting, and a function modeling the angular color variations. The color is then prefiltered with a kernel defined by the environment map and brightness BRDF. Marginalizing over the azimuthal rotation around the ideal reflection direction results in a 3D prefiltered color table, parameterized by the ideal reflection direction and the angle between viewing direction and surface normal, which can further be decomposed into two functions depending on one of the two parameters, respectively. The additional data that needs to be stored with our method consists of a low-resolution cubemap and 2D array per material and environment map combination.