[Bf-funboard] Volumetric normals based on density gradient?

[Tiago Estill de Noronha]

Just 4 points might not produce the most accurate result, hence why i said
N; make it an user settable number so the user can choose how much time
they want to exchange for accuracy.

I see this being useful for all sorts of complex materials; liquids with
bubbles, transparent stuff with non-homogeneous density (like mixed
temperature air, some types of glass, gravitational lensing and so on),
those fancy stones that got reflective iridescent flakes inside etc


On Fri, Apr 4, 2014 at 6:55 AM, Brecht Van Lommel <
brechtvanlommel at pandora.be> wrote:

> Volume scattering functions don't use normals, so I guess you want to get
> surface BSDFs working for volumes, which then need a normal?
>
> Computing the gradient would make the material evaluation maybe 4x slower,
> comparing to a volume scatter material, due to evaluating the texture at 4
> points. The overall render time impact may not be that bad in practice
> though.
>
> On Apr 4, 2014 12:25 AM, "Tiago Estill de Noronha" <
> TiagoTiagoT+Bf-funboard at gmail.com> wrote:
> >
> > How big would the impact in the performance be if the normals for
> > volumetric materials were calculated based on the density gradient (i
> guess
> > sampling N points around the point in question to see which direction is
> > denser; averaging the directions in case there was disagreement)?
> >
> >
> >
> > And for homogenous density regions perhaps it could do something like
> > produce a <0,0,0> normal internally that gets treated as a sorta
> wildcard,
> > pointing back at the direction of incoming rays?
> > _______________________________________________
> > Bf-funboard mailing list
> > Bf-funboard at blender.org
> > http://lists.blender.org/mailman/listinfo/bf-funboard
> .
> _______________________________________________
> Bf-funboard mailing list
> Bf-funboard at blender.org
> http://lists.blender.org/mailman/listinfo/bf-funboard
>