Shape-from-polarisation is a method for determining the orientation of
surfaces based on the polarisation state of light reflected from
the surface.
In [Atkinson and Hancock 2006a]
we showed how Fresnel theory can be used to relate the
angle of reflection to the polarisation state of light in the
diffuse reflection regime. This allows the recovery of surface
normals from polarisation images of objects, although the directions are still ambiguous.
These ambiguities can be resolved using photometric stereo[Atkinson and Hancock 2005a]
We extended this work in [Atkinson and Hancock 2005b]
where we established a directional reflectance model for the surface.
Stereo
algorithms can be used to extract 3D structure from a pair (or more) of
images. Stereo is very effective where good correspondences can be
found, i.e. where there is clear surface texture, but fails on smooth
surfaces. Surface orientation can be found from a variety of sources,
such as shape from shading or texture, and can be used to extract
3D information
from a single image. However, this information is not accurate enough
or complete enough to satisfactorily reconstruct the surface. The
aim of this
work is to combine these two methods to produce a more effective and
complete reconstruction. The combination can be done through a
probabilistic model of the surface depth, such as a suitable Markov
random field.