Pont and Koenderink discuss the complexity of reflectance in real-world circumstances. Current reflectance models assume glossy surfaces, and, consequently, are able to represent their reflectance by a linear combination of specular and diffuse reflectance functions.
The paper starts with a thorough overview of research done on reflectance distribution functions. Next, the authors present their photographs of three physical objects: one painted with matte white paint, one painted with glossy white paint, and a polished black plastic model. Since diffuse illumination was used, a direct comparison was possible. A visual analysis was done to obtain some intuitive understanding of the problems present with the reflectance of rough surfaces. To determine the influence of the differences in texture between the models, a semi-quantitative analysis was conducted.
The authors start with these assumptions: material scatters light via body scattering and surface scattering, and, for each type of scattering, there is a certain probability that each type of scattering will occur. The total reflectance is then a linear combination of the latter probabilities.
Using an approximation of the reflectance of objects, the authors were able to calculate the radiant power of first- and higher-order reflections of each of the models. Their results demonstrated that linear superpositions of textures of locally specular pitted surfaces, and of locally diffuse pitted surfaces, differ substantially from the texture of glossy pitted surfaces. Realistic reflectance models should take the photometrical effect of multiple reflections into account, except in the case of perfectly smooth convex dielectrics.