Reflectance Models | Radiometry and Reflectance

This lecture from the ‘First Principles of Computer Vision’ series, presented by Shree Nayar, introduces fundamental reflectance models used in computer vision. It begins by distinguishing two primary reflection mechanisms: surface (specular) reflection, which occurs at the interface and produces glossy appearances, and body (diffuse) reflection, where light penetrates the material and scatters internally, resulting in matte appearances. The lecture then formalizes these concepts using the BRDF (Bidirectional Reflectance Distribution Function). The Lambertian model, proposed by Lambert in 1760, describes ideal diffuse reflection where surface radiance is constant in all directions, leading to a BRDF that is a constant (albedo divided by pi). The lecture derives the relationship between radiance and irradiance for Lambertian surfaces, showing that radiance depends on the cosine of the incidence angle. The ideal specular model, at the other extreme, describes mirror-like reflection where all incident light is reflected in a single direction, the specular direction, which is the reflection of the source direction about the surface normal. The BRDF for ideal specular reflection is expressed using delta functions. The lecture illustrates these models with examples, such as a clay vase (mostly body reflection) and a shiny object (surface reflection), and discusses hybrid reflection combining both. It also visually demonstrates how a Lambertian sphere appears with varying brightness based on incidence angle, and how an ideal specular sphere shows a single bright point. The lecture concludes by noting that these simple models are widely used due to their simplicity and ability to approximate many real-world surfaces.