Improved Diffuse Reflection Models for Computer Vision

  • Authors:

  • Lawrence B. Wolff;Shree K. Nayar;Michael Oren

  • Affiliations:

  • Computer Vision Laboratory, Department of Computer Science, The Johns Hopkins University, Baltimore, MD 21218;Center for Research in Intelligent Systems, Department of Computer Science, Columbia University, New York, NY 10027;Artificial Intelligence Laboratory, Massachusetts Institute of Technology, Cambridge, MA 02139

  • Venue:
  • International Journal of Computer Vision
  • Year:
  • 1998

Quantified Score

Hi-index 0.00

Visualization

Abstract

There are many computational vision techniquesthat fundamentally rely upon assumptions about the nature ofdiffuse reflection from object surfaces consisting of commonlyoccurring nonmetallic materials. Probably the most prevalentassumption made about diffuse reflection by computer visionresearchers is that its reflected radiance distribution isdescribed by the Lambertian model, whether the surface is roughor smooth. While computationally and mathematically arelatively simple model, in physical reality the Lambertianmodel is deficient in accurately describing the reflectedradiance distribution for both rough and smooth nonmetallicsurfaces. Recently, in computer vision diffuse reflectancemodels have been proposed separately for rough, and, smoothnonconducting dielectric surfaces each of these modelsaccurately predicting salient non-Lambertian phenomena that haveimportant bearing on computer vision methods relying uponassumptions about diffuse reflection. Together thesereflectance models are complementary in their respectiveapplicability to rough and smooth surfaces. A unified treatmentis presented here detailing important deviations from Lambertianbehavior for both rough and smooth surfaces. Somespeculation is given as to how these separate diffusereflectance models may be combined.