Help Guide Disclaimer Contact us Login
  Advanced SearchBrowse





Surface reflectance estimation under unknown natural illumination


Fleming,  RW
Research Group Computational Vision and Neuroscience, Max Planck Institute for Biological Cybernetics, Max Planck Society;
Department Human Perception, Cognition and Action, Max Planck Institute for Biological Cybernetics, Max Planck Society;

There are no locators available
Fulltext (public)
There are no public fulltexts available
Supplementary Material (public)
There is no public supplementary material available

Fleming, R., Dror, R., & Adelson, E. (2001). Surface reflectance estimation under unknown natural illumination. Poster presented at First Annual Meeting of the Vision Sciences Society (VSS 2001), Sarasota, FL, USA.

Cite as:
Purpose : Physical surfaces such as metal, plastic, or paper have different optical qualities that lead to different characteristics in images. The actual image of a surface, however, depends heavily on the illumination. This experiment measures the accuracy with which humans match surfaces of similar reflectance properties under different natural illuminations, in the absence of contextual information to specify illumination. Methods : A set of balls were computer rendered as if illuminated by 9 real-world 360° illumination maps, taken from Debevec et al. (2000). The balls' reflectance properties were parameterized using Pellacini et al.'s (2000) psychophysically uniform reparameterization of the Ward reflectance model. A range of values determined the magnitude of the specular lobe, c, and the sharpness of the specular lobe, d. Subjects were presented with 2 balls of randomly selected c and d values rendered with different illumination maps. They then had to adjust c and d for one of the balls (the ‘match’) until its perceived reflectance qualities were the same as for the other ball (the ‘test’). All balls were viewed against a black background, in the absence of contextual information. Results : Correlation and regression analyses reveal that subjects perform accurately and reliably for both c (slope=0.83; correlation coefficient=0.79) and d (slope=0.90; correlation coefficient=0.90). Furthermore they performed significantly better than with a set of contrast-polarity inverted images derived from the original set, which produce no coherent percept of surface reflectance but which contain spatially identical image properties (p<0.05). Conclusions : Contextual information is not required for reliable matches of material qualities.