Modeling and Verifying the Polarizing Reflectance of Real-World Metallic Surfaces

To generate convincingly rendered images of surface reflection appearances, computer graphics usually relies on bidirectional reflectance

distribution function (BRDF) models. One such model is the Phong reflectance model (see the “Related Work” sidebar). In believable rendering applications, these intuitive models don’t have to be physically accurate as long as the resulting image satisfies the viewer’s expectations. For example, the traditional Phong model is not energy conserving. However, predictive rendering, which we focus on here, relies on BRDF models to make valid assumptions of the surface’s physical

behavior to correctly describe the light transport in complex scenes. Two well-known examples of physically plausible BRDF models are the Fresnel reflectance and Torrance-Sparrow models (see the “Related Work” sidebar). They describe the surface reflectance behavior down to its reflective polarization properties. To verify the accuracy of these physically based models, we collected a set of metallic surfaces. Then,

we used an ellipsometer to measure the polarization properties of their reflected light. Finally, we compared the measurements to the reflectances we predicted from the physically plausible BRDF models.