Margus Ramst <mar### [at] peakeduee> writes:

> Thomas Willhalm wrote:
> > 
> > (Please excuse my bad English)
> > In the current picture, the lighting model isn't correct. Consider a
> > point P on the surface of a hair. The normal vector of the hair at P
> > is _not_ used to determine the amount of light that travels from a light
> > source to the eye. At the moment, only the directions from P to the light
> > source and from P to the eye are used.
> > 
> > Furthermore, I would like to vary the colour of the fur.
> > 
> > Simply speaking, I want to apply a texture to a media.
> > 
> 
> I kind of doubt you can apply the phong (or whatever) lighting model to
> media, since media does not have surface normals. It is a volume effect and
> doesn't even have a surface as such. But perhaps you know better.

As already mentioned, I try to implement the method described by Perlin
and Hoffert. They used the gradient of the density function to simulate
a surface normal. The images that they have in their paper look much
more realistic than my results so far.

> If you use the superpatch you _might_ achieve the fur effect with a
> isosurface, since you have apparently developed a function that gives this
> hairy appearance. If so, you wouldn't need to worry about applying lighting
> models and textures.

First, I want to point out that it's not "my" function, but the function
that is described in the paper. I have tried to use it as isosurface, but
the outcome wasn't very realistic. The clue with this function is that
it is used with partial transparent media. (At least, this is what I 
think now.)

> Oh, and about your furball - it really does look like fur. A bit
> transparent, but that should be easy to fix. Can it also make hair at
> different angles? 

I'm not totally sure about what you mean. You can't use this method 
to create single hairs that are next to each other but point in different
directions (or more precisely: I don't know how to do it). neighboured
hairs have to be almost parallel.

You can however add some turbulence to the function. This can result
in hairs that point in different direction at different regions of the
workpiece. Of course, you could also use a non-random function to
change the direction.

> And how long does it take to render?

Between 8 and 17 minutes with a "336 MHz SUNW,UltraSPARC-II" CPU.
The memory consumption is quite low. POV-Ray reports about 100 KB.

Thomas


-- 
http://www.fmi.uni-konstanz.de/~willhalm

Post a reply to this message