Robust Estimation of Reflectance Functions from Polarization.

This paper presents a new approach to image-based reflectance function estimation. The method first uses Fresnel theory for reflection and polarization analysis to estimate the surface orientation for each image pixel from a single view. The method is confined to the case where the light source and camera lie in the same direction from the target. A 2D histogram of surface zenith angles and pixel intensities is then calculated. This histogram is processed using a robust and computationally efficient statistical analysis. Histogram data are fitted to probability density functions to deduce the reflectance function. Objects of varying complexity and material are analysed and compared to ground truth. [ABSTRACT FROM AUTHOR]