Advanced Search

Browse by Discipline

Scientific Societies

E-print Alerts

Add E-prints

E-print Network

  Advanced Search  

Specular highlights of plastic surfaces and the Fresnel Coefficient Elli Angelopoulou*

Summary: Specular highlights of plastic surfaces and the Fresnel Coefficient
Elli Angelopoulou*
and Sofya Poger

Computer Science Department, Stevens Institute of Technology, Hoboken, NJ 07030, USA
Computer Science Department, Iona College, New Rochelle, NY 10801, USA
One of the biggest clues in specularity detection algorithms is the color of the specular highlights. There is a prevalent
assumption that the color of specularities for materials like plastics and ceramics can be approximated by the color of the
incident light. We show that such an assumption is not generally appropriate because of the effects of the Fresnel reflec-
tance coefficient and its dependence on wavelength. Our theoretical analysis establishes that the sensitivity of the Fresnel
term to the wavelength variations of the refractive index can be at least as large as 15%. Our experiments demonstrate
that, even with traditional RGB color cameras, the recorded color of specular highlights is distinct from the color of the
incident light. Furthermore, by computing the spectral gradients (i.e. the partial derivatives of the image with respect to
wavelength) at specular regions we can isolate the Fresnel term up to an additive illumination constant. Our theory is sup-
ported by experiments performed on multispectral images of different colored plastic tiles. The refractive indices of the
opaque plastics were measured using a specialized spectroscopic ellipsometer. The computed spectral gradients of the tile
specularities exhibited a less than 2.5% deviation from the predicted theoretical values.
Keywords: specular highlights, plastics, Fresnel term, index of refraction, spectral gradients.


Source: Angelopoulou, Elli - Department of Computer Science, Friedrich Alexander University Erlangen Nürnberg


Collections: Computer Technologies and Information Sciences