Separation of Reflection Components Using Color and Polarization

  • Authors:

  • Shree K. Nayar;Xi-Sheng Fang;Terrance Boult

  • Affiliations:

  • Department of Computer Science, Columbia University, New York, N.Y. 10027/ E-mail: nayar@cs.columbia.edu, fang@cs.columbia.edu, boult@cs.columbia.edu;Department of Computer Science, Columbia University, New York, N.Y. 10027/ E-mail: nayar@cs.columbia.edu, fang@cs.columbia.edu, boult@cs.columbia.edu;Department of Computer Science, Columbia University, New York, N.Y. 10027/ E-mail: nayar@cs.columbia.edu, fang@cs.columbia.edu, boult@cs.columbia.edu

  • Venue:
  • International Journal of Computer Vision
  • Year:
  • 1997

Quantified Score

Hi-index 0.00

Visualization

Abstract

Specular reflections and interreflections produce strong highlights in brightness images. These highlights can cause visionalgorithms for segmentation, shape from shading, binocular stereo, andmotion estimation to produce erroneous results. A technique is developedfor separating the specular and diffuse components of reflection fromimages. The approach is to use color and polarization information,simultaneously, to obtain constraints on the reflection components ateach image point. Polarization yields local and independent estimates ofthe color of specular reflection. The result is a linear subspace incolor space in which the local diffuse component must lie. This subspaceconstraint is applied to neighboring image points to determine thediffuse component. In contrast to previous separation algorithms, theproposed method can handle highlights on surfaces with substantial texture, smoothly varying diffuse reflectance, and varying materialproperties. The separation algorithm is applied to several complex sceneswith textured objects and strong interreflections. The separation resultsare then used to solve three problems pertinent to visual perception;determining illumination color, estimating illumination direction, andshape recovery.