[Bf-committers] Volumetrics Info

[GSR]

Hi,
echelon at infomail.upr.edu.cu (2008-04-16 at 1149.47 -0400):
> > Mirror spheres? What about two opposing mirrors as used for infinite
> > tunnel SFX? Or 20 layers of glass?
> Yes, I know that when the reflection componet is on, raydeph count on
> render times (a lot), I was talking to spheres without reflections.

With refractions? If no reflections or refractions, I do not see why
it would increase time, as there is no recursivity (AO or ray lamps
would be just one extra level, IIRC, so constant).

> > BTW, looking at your images, I see they suffer some serious banding,
> > any idea about how to workaround that?
> Im also know the banding effect, the raycaster is far from completed and
> for that reason i haven't made yet testing builds, in these cases I'll
> make also a tresholded approach so is up to the user how much
> quality/speed he could afford. (there are volumetrics setups with low
> presition that do not show any bandings and are faster and other will
> require filterings and igher presition)

There are methods that give better looking results without wasting
time in more samples. For example generate a grid (16*16, more?) of
small (0.0 to 0.something) pseudo random numbers, and offset the
sampling depth in a per pixel basis. You can find papers talking about
this dithering trick from around the 90s, some even did it in
hardware.

GSR