Spectral photon mapping

Abstract

This paper introduces a photon mapping framework that accommodates wavelength-dependent refraction through a model based on the Sellmeier equation. In contrast to the standard RGB-based techniques, the new method traces photons through several spectral channels to better represent dispersion and color shift within transparent media called 'spectral bins'. During the first stage, the photons are released from light sources and scattered arbitrarily throughout the scene, each path being reflected, refracted, or absorbed according to wavelength-dependent characteristics. Afterward, the photons are separated into two K-D trees that differentiate between refractive and regular interactions. In a second pass, a reverse ray-tracing technique approximates final pixel intensities by fetching local photon density estimates for diffuse lighting and recursively systematically calculating specular reflections and refractions.

The model supports a realistic material dispersion of materials such as glass and water by calculating refractive indices at specific wavelengths. One can apply the Sellmeier equation, enabling significant color shifting and multicolored caustics within transparent objects. Russian roulette sampling also eliminates the issue of very long paths while maintaining statistical accuracy, and parallelization alleviates computational overhead. The experimental findings validate that the spectral method effectively reproduces intricate optical phenomena, such as multicolored caustics, color fringing, and secondary color bleeding, thus providing a step up from conventional three-channel models in physical realism. The findings demonstrate the applicability and versatility of spectral photon mapping to high-quality global illumination and optical simulation.