Robust and accurate estimate of the orientation of partially polarized light from a camera sensor

Polarized light carries valuable information about where the light has been and the various physical parameters that have been acting upon it. Thus th ere are several methods in computer vision that make it possible to obtain information on the observed object by studying the polarization of light re flected on the object. Most studies using this principle are interested in the determination of the object orientation in three-dimensional space. The basis of these studies rests on the estimate of a parameter that connects the orientation of the observed surface and the polarization of the reflect ed light wave. This parameter is the angle of polarization phi, also called the orientation of polarization. Generally, one using these methods estima tes the phi angle by observing the reflected light wave through a linear po larizing filter and grabbing multiple frames for different angular orientat ions of the polarizer. So, between each acquisition, the polarizer is rotat ed of an angle theta relative to a horizontal reference axis. The accuracy of the phi estimate is then directly related to the positioning of the pola rizer. But, in practice, it is difficult to guarantee the exact value of th e rotation of this polarizer. It is all the more difficult to guarantee the reliability of positioning in time. We thus propose a robust and accurate solution, based on the self-calibration principle, for measuring the orient ation of partially polarized light using CCD cameras. In contrast to method s generally discussed in computer vision journals, our estimate of the angl e of polarization is independent of the reliability of the polarizer positi oning. (C) 2001 Optical Society of America.