genBRDF: Synthesizing Novel Analytic BRDFs with Genetic Programming

We present a framework for automatically synthesizing new analytical BRDFs to model classes of measured materials using a heuristic-based search commonly called Genetic Programming. Our approach can be seen as an extension of traditional approaches to reflectance modeling that rely either on a phenomenological process or a physically-based model. In both cases, our technique augments the human effort involved in deriving suitable mathematical approximations of complex reflectance functions with a large-scale search. We present results that show our technique is able to find novel mathematical expressions that are adjustable, compact, and more accurate than previously available alternatives for representing common classes of materials like metals and dielectrics. We also describe a new fast approximation of a common perceptual BRDF error metric used to make this search tractable. CR Categories: I.3.3 [Computer Graphics]: Three-Dimensional Graphics and Realism—Display Algorithms I.3.7 [Computer Graphics]: Three-Dimensional Graphics and Realism—Radiosity;