Colour and Reflectance
Dr. Graham D. Finlayson
Department of Computer Science,
The University of York,
York YO1 5DD.
United Kingdom
Abstract
In order for computer
vision to progress we need physically accurate models of
surface reflection. Indeed intrinsic scene
properties such as surface colour, surface shape and surface texture
can only be recovered
when surface reflection is understood. Of course
an accurate model while necessary for estimation, is not in itself sufficient.
Even when the principles of reflection are understood (e.g. when surfaces
adhere to the well-known Lambertian
or dichromatic models) scene properties have proven to be hard
to extract.
In this report I outline some of the current (and future) research directions in the measurement
of surface reflection, the
modelling of surface reflection and
the estimation of reflection parameters
(including surface spectral reflectance, surface colour
and reflectance invariants).
As a subtext this report advocates
the adoption of colour cameras as the standard tool in computer
vision.