Christoph Hormann wrote:

> Wolfgang Wieser wrote:
>> I am not completely sure about what I am talking here but at least
>> what I propose in the following lines removed black spots on the
>> isosurface of my Mars renderings. (These spots only occur when the light
>> comes under a flat angle and when no_shadow is NOT set.)
>> 
>> [...]
> 
> What you describe seems to be specific to shadow rays, does your
> suggested change also influence the appearance of the actual surface
> (i.e. camera rays). You could try rendering with a fairly large accuracy
> value for testing.
> 
Well, in iso code there is no difference between viewing and shadow rays. 
The in_shadow_test flag is never actually used except for some cache 
which is handeled at a higher level and should not matter here. 

What I describe seems to be specific to rays which come in a _very_ 
flat angle over the surface. And for these Mars renderings, the 
shadow rays come really flat on the shadow side of hills. 

> I agree with Thorsten that the interpolation will probably cause
> slowdown without much advantage 
>
I am not sure about that. Because the gain in accuracy is considerably. 
The original code will track the ray until its position is known to 
be within "surface +/- accuracy". But with the modification I propose, 
we gain significantly in accuracy because we can linarly interpolate the 
"exact" location of the surface. 

> (and i don't think the simple version 
> you posted will work correctly in all situations).  
>
You don't _think_ ...
Okay, I am not completely sure, too. 
Actually, 
--------------------------
    if(EP2->f<=0.0)
    {
            double df = EP1->f-EP2->f;
            // Need to calc (EP1->t*EP2->f - EP2->t*EP1->f) / df
            // in a numerically stable way. 
            ISOSRF->tl = (df>1e-14) ? 
            EP2->t + t21*EP2->f/df : 
            0.5*(EP2->t+EP1->t);
        return true;
    }
    return false;
--------------------------
should work as well (and does for my scenes). All the other conditions 
are just to work around conditions which I am not sure about. 

> Also note we are 
> talking about differences of a maximum of accuracy*0.5.
> 
So what?

Please see the following web page for a comparison: 
http://www.cip.physik.uni-muenchen.de/~wwieser/tmp/poviso.html

Wolfgang

Post a reply to this message