[Radiance-general] bean counting the photon array

[Chris Kallie]

Fellow Renderers,

I am using The Radiance Photon Map Extension to analyze the light 
passing through an irregular glass object. The luminaire is not visible 
from the viewpoint, and the environment is a simple Lambertian surround.

I wish to account for all scattering events (i.e., specular reflections 
and refractions) that eventually make it to the viewpoint, and I wish to 
render (or at least account for) scattering events incrementally, in 
order to visualize the unfolding (and hopefully the order) of 
contributing events. I have been successful (to some degree) when the 
environment has a non-zero reflectance, but mkpmap fails when I set the 
surrounding environment reflectivity to zero.

So far, the only way I have been able to record a single bounce in a 
non-reflective environment is with rpict, but there I notice not just 
one, but also two or three scattering events: 1) an expected specular 
reflection on the outer hull of the object; plus 2|3) what appears to be 
an inner specular reflection after an initial refraction--followed by an 
exit refraction--from within the object. This indicates to me that rpict 
does not count refractions the same way it counts specular reflections 
(although I think mkpmap does). I have also been able to render 1+ 
scattering event(s) with mkpmap, plus 1+ scattering event(s) with rpict, 
when including a non-zero reflectance Lambertian environment, but I 
don't think those images are useful because I don't think they describe 
any realistic physics (although I might be able to use similar images 
with pcomb in some clever way to reverse-engineer the contributions of 
physical events).

The other issue I have with using rpict for computing the first 
scattering event is because I know it under-samples the caustic 
environment without mkpmap, so I am guessing that rpict will be useful 
for computing some proportion of all but (at least) the first scattering 
event(s) (i.e., perhaps I could use some thoughtful combination of 
n+p-scattering events between n-scatter-distributions and 
p-scatter-gatherings).

I hope this description/shorthand wasn't too confusing.

In summary, I would like to use a combination of mkpmap and rpict to 
incrementally build the photon array, and perhaps use pcomb or some 
other statistical or combinatorial technique to analyze the 
contributions of scattering events as they relate to the viewpoint. 
Finally, whatever limitations are imposed by the software can be 
incorporated into the limits of the model (in other words, I'll take 
what I can get!).

P.S. Unrelated question: How do I tell Radiance the right physical units 
for my glass object?

Thank you, always!
Chris