Comparison of polarization bidirectional reflectance distribution function (BRDF) models

IR target signature models rely on some representation of the BRDF to describe surface optical properties. For unpolarized signature modeling, it usually suffices to split the BRDF into a diffuse part and a specular part, and to model each part with a few parameters which fit the data; the Sandford-Robertson approach is perhaps the best-known. Artifacts produced by this idealization are almost always small compared with true signature features. For polarization modeling, however, one computes quantities, such as the degree of polarization, which are small differences of large numbers, and which derive from small differences between elements in the representation of the polarization BRDF. An adequate representation is difficult to determine because the conventional scalar BRDF function generalizes to a matrix function with typically 6-10 independent elements, requiring that numerous parameters be determined from data in a consistent manner. This requires a new BRDF representation which is based on physical principles, and which is comprehensive and accurate enough to satisfy requirements for use in signature modeling. This paper considers several candidate schemes for parameterizing the polarization BRDF from the standpoints of consistency and the ability to model all important features of the surface optical properties, without introducing so many parameters as to be ill-conditioned