Tales from the coal face

global illumination and geometry problems within Computer Graphics,
I have encountered a number of difficult problems that are not
commonly addressed within the academic rendering community.

The talk will cover problems that can often be overcome with
pure computing power, but only by using prohibitive amounts of it
- intelligent alternatives are sought. Many of the issues only
emerge when software developed during research is extended, to cope
with the rigours and demands of commercial software. Several example
problems are covered in some detail.

Non-Lambertian light sources, daylighting, light sources made up from
thousands of polygons, coloured light sources, caustics effects with
light-forwards ray tracing, and refinement oracles for radiosity,
all play a part in the talk.


In many ways, the talk is an appeal for co-operation; why are such
problems ignored by the rendering community? Should research be
directed towards the solution of such problems, or are they truly
the problem of companies that wish to benefit from the work of
academics?