A Compression Method for Spectral Photon Mapping

Lai,G., Christensen,N.J.

Abstract:
The photon map method can easily be extended to handle wavelength-dependent phenomena such as dispersion, chromatic aberration, etc. Using the computationally effective approach of point sampling for this extension, the size of the photon map is increased proportionally with the number of samples. In this paper we present a solution for modelling wavelength-dependent phenomena that keeps variance low, while having a memory usage comparable to that of an RGB based renderer. The method is best used for photon mapping, where there is a need to store large amounts of spectral flux directly in the photon map.
Our method incurs a slight loss of accuracy for photons in the global map, while photons in the caustics map retain all information. Our tests show negligible loss of accuracy in the image quality.