Polarimetric Effects in Non-polarimetric Imaging

Radiative transfer is commonly modeled as the propagation of unpolarized radiation. More accurate approaches utilizing polarimetric quantities are usually only applied to sensors that purposefully discriminate polarimetric information. In this paper, we examine the effect on fidelity of utilizing polarimetric radiative transfer modeling for non-polarimetric sensors. We show that if the primary irradiance on a scene is significantly polarized (such as sky irradiance) then polarimetric radiative transfer modeling is warranted and provides a significant increase in fidelity. We demonstrate this effect by performing target detection of shadowed man-made objects in real and simulated imagery.