The bidirectional reflectance distribution function (BRDF) describes material reflectance by relating incident irradiance to scattered radiance. One popular class of BRDF models is the microfacet model, which assumes geometric optics but is more readily applicable to remote sensing. One drawback of this geometric optics model is the need for a cross section conversion term, which diverges at grazing angles. This problem is only partially addressed by adding a geometric attenuation term to conserve energy, while still neglecting wave optics effects. Based on previous work comparing microfacet and wave optics models, Butler proposed to replace the geometric attenuation and cross-section conversion terms with a theoretical approximation, the closed-form polarization factor, Q. Analysis presented both at Optics and Photonics by Butler in 2017 and SPIE Defense and Commercial Sensing (DCS) by Ewing in 2018 show this modification to be effective for both high density (but low fidelity) data, and low density (but high fidelity) data, particularly at grazing angles, but that analysis only examined unpolarized data. In this work, the theoretical modification is analyzed using high fidelity, low density, in-plane polarimetric oblique and grazing angle BRDF data. These polarimetric data are fit to the novel version of the microfacet model for each polarization separately, using the polarization factor Q, and the error in the fits are compared to the unpolarized fits that were presented at SPIE DCS. These results suggest incorporating the polarization factor to improve the quality of fit consistently for materials, including substantial improvement at grazing angles.
|