CVA Special Issue: Surface Detail in Computer Models

Traditionally in vision, the reflectance of light at a surface has been modelled with rather simple linear combination of Lambertian and specular models. Such models are applicable for manmade scenes and where the surfaces are either “matte” or shiny, and rather uniform. An office scene with a desk, chalk, pen, stapler may be modeled well with these simple models. But real world scenes have a far more complex interaction of light and surfaces. Consider Figure 1 which shows examples of real-world objects and their fine scale surface appearance. As high resolution imagery becomes more prevalent, these fine scale details are relevant. Note that these details are not simply painted patterns on a surface, rather the material properties and the fine scale geometry cause a complex interaction of light at the surface.This means that it’s not trivial to predict the appearance of the surface when the light source and camera moves. So why are these details important? In computer graphics, we want to render realistically so computationally modelling the appearance of surface under variations of viewing and illumination direction is required. In computer vision, surface details provide an important cue for recognition. Using these fine scale details to assist recognition is