Temporal radiance caching

  • Authors:
  • Pascal Gautron

  • Affiliations:
  • Thomson Corporate Research, France

  • Venue:
  • ACM SIGGRAPH 2008 classes
  • Year:
  • 2008

Quantified Score

Hi-index 0.00

Visualization

Abstract

Very often global illumination methods aim at simulating the light/matter interactions within static scenes. Accounting for the displacement of objects and light sources either require a complete recomputation of the global illumination solution for each frame of the animation, or involve complex data structures and algorithms for temporal optimization. Furthermore, the global illumination solutions commonly exhibit low temporal quality when used in dynamic scenes: flickering, popping, ... In the context of computer-assisted effects for movies, high quality global illumination is obtained through temporal filtering: a 30 fps animation is first rendered at 60 fps by recomputing the global illumination for each frame. Then, each frame of the 30 fps animation is generated by averaging two frames of the 60 fps animation, hence reducing the temporal artifacts at the cost of high computational cost. For interactive applications such as video games, the illumination must be computed interactively. In this case, approximate models are generally preferred, such as the precomputation of a static global illumination solution, and the update of direct lighting only at runtime. This chapter describes a simple and accurate method based on temporal caching for the computation of global illumination effects in animated environments [GBP07], where viewer, objects and light sources move. This approach focuses on a temporal optimization for lighting computation based on the irradiance caching [WRC88] technique. As this algorithm leverages the spatial coherence of indirect lighting to reduce the cost of global illumination, we consider here an extension of these methods for sparse temporal sampling and interpolation. In [WRC88], Ward et al. propose a reuse an irradiance value in the neighborhood of the actual computation point. While the weighting function and gradients of [WRC88] account for the spatial change of irradiance, Temporal Radiance Caching considers the temporal change of the indirect lighting (Figure 8.2).