Nicolas Alvarez <nic### [at] gmailisthebestcom> wrote:

> What happens if two spheres intersect, but none is completely inside the
> other? Which sample count remains?

This is where things get tricky (and is a notorious example of one's sphere's
outline showing up in the other.) The "overlapping" parts of
the two spheres' medias is determined by the samples count of the "larger" of
the two...and/or by which sphere the camera ray hits first, I believe. Visible
sphere outlines arise when samples in the two spheres are unequal (and just too
low); making the samples equal--though not necessarily high--usually eliminates
that. ("Equal" depends on the relative sizes of the spheres, though.) The parts
of the spheres that are NOT intersecting take on their own individual samples
count...though the camera position may have something to do with that
as well.

The worst-case scenario seems to be when the two spheres intersect and the
intersection surfaces are seen "edge on" so to speak--what looks like a convex
lens shape. I've had to really pump up (equal) samples counts in the two to fix
that; just increasing one or the other won't do it.

Ken W.

Post a reply to this message