Color constancy for stereo imaging

Date of Award


Degree Name

M.S. in Electrical Engineering


Department of Electrical and Computer Engineering


Advisor: Keigo Hirakawa


Reflectance values of real-world objects are better approximated by Bidirectional Reflectance Distribution Function (BRDF) models than Lambertian models. In this thesis, we propose a 3D color constancy method to take advantage of the angular dependencies of reflectance values. Stereo matching techniques are used to reconstruct the left camera image form the right, then the difference between the original left camera image and the reconstructed one is assumed to be caused by the BRDF. The illuminant is estimated by fitting a line to this difference image in the total least squares sense. This is in contrast to the existing color constancy methods which use a single 2D image to estimate the illuminant. Experimental results show that 3D color constancy offers a significant improvement over the 2D methods.


Stereoscopic cameras, Reflectance, Image reconstruction

Rights Statement

Copyright © 2012, author