All-frequency relighting of non-diffuse objects using separable BRDF approximation

  • Authors:
  • Rui Wang;John Tran;David Luebke

  • Affiliations:
  • Department of Computer Science, University of Virginia;Department of Computer Science, University of Virginia;Department of Computer Science, University of Virginia

  • Venue:
  • EGSR'04 Proceedings of the Fifteenth Eurographics conference on Rendering Techniques
  • Year:
  • 2004

Quantified Score

Hi-index 0.00

Visualization

Abstract

This paper presents a technique, based on pre-computed light transport and separable BRDF approximation, for interactive rendering of non-diffuse objects under all-frequency environment illumination. Existing techniques using spherical harmonics to represent environment maps and transport functions are limited to low-frequency light transport effects. Non-linear wavelet lighting approximation is able to capture all-frequency illumination and shadows for geometry relighting, but interactive rendering is currently limited to diffuse objects. Our work extends the wavelet-based approach to relighting of non-diffuse objects. We factorize the BRDF using separable decomposition and keep only a few low-order approximation terms, each consisting of a 2D light map paired with a 2D view map. We then pre-compute light transport matrices corresponding to each BRDF light map, and compress the data with a non-linear wavelet approximation. We use modern graphics hardware to accelerate precomputation. At run-time, a sparse light vector is multiplied by the sparse transport matrix at each vertex, and the results are further combined with texture lookups of the view direction into the BRDF view maps to produce view-dependent color. Using our technique, we demonstrate rendering of objects with several non-diffuse BRDFs under all-frequency, dynamic environment lighting at interactive rates.