ivfer wrote:
> 
> 1) Did you ever solve the problem with splotchiness when firing random
> photons? If you're still using a scanning approach do you handle the case
> where an object surrounds a photon source? You'd need to scan based on
> spherical coordinates in that case.

No.  I use a jittered uniform scan.  And I always use spherical coordinates.
That way it provides even distribution of photons on the surface of a
sphere centered on the origin, which is more physically accurate (as
opposed to evenly distributed on a planar surface).

> 2) Are you using wavelength dependent ior or are the r,g,b caustics due to
> different positions of the lights? If you're doing wl-dependent ior, how do
> you encode the light's spectrum in the photon? I punted and simply fired
> multiple single-wavelength photons (jittered spectrum sampling) if the
> target had a wl-dependent ior (and prayed that subsequent objects weren't
> wl-dependent).

The rgb caustics come from three lights in different positions.  This does
do wavelength dependent ior, too (images attached).  Instead of sampling
the spectrum randomly, I allow the user to give a list of wavelengths
(in the form of RGB vectors) that make up the light.  I use a modified
version of Daren Wilson's dispersion patch, along with your rgb->hue
conversion function to do the rest.

-Nathan

Post a reply to this message

Attachments:
Download 'prism.jpg' (16 KB) Download 'prism2.jpg' (8 KB)

Preview of image 'prism.jpg'

Preview of image 'prism2.jpg'