Applications Using Photometric Stereo (PS)

Classifying rough surfaces using reflectance

The original image (left) and its topographical component (right).

The aim of this work is to discover as much as we can about a sample without actually touching it. In many applications, e.g. forensics, we are prepared to go to some trouble to extract as much information about a surface as possible. Because a sample's reflectance properties are a function of the material, or materials, of which it is composed, reflectance is a powerful, physically-based, cue for classification. The image we observe is a function of both surface topography and reflectance. We must first resolve these two components. Figure above shows an image of a (glossy) ammonite embedded in a (diffuse) matrix: the original image (left) contains both reflectance and topographical information; the image on the right shows our estimate of the component of the image due to surface topography. This work uses a novel variant of photometric stereo to simultaneousely estimate both the shape and the reflectance characteristics of the sample.