Rafal 'Raf256' Maj schrieb:

> Tim Nikias <tim### [at] gmxde> wrote in news:3D29C26F.26C123D6@gmx.de:
>
> > Radiosity, AFAIR, doesn't affect phong and specular. Those are
> > lightsource dependant "faking" algorithms.
> >
> > If I'm correct, specular and phong only take the surface-normal
> > in relation to the direction towards the lightsource and calculate
> > a "shinyness" based on that angle. Since radiosity isn't connected
> > to a lightsource, the implementation of light-directions from radiosity
> > is fairly difficult.
>
> but IMHO this is worth the effort :)
>
> afair realy radiosity needs calculations of about 50-200 (count)
> samples per one pixel. Due to reuse - only few samples are calculated -
> because POV remebers radiosity brighntess in last pixel(s)... em I right ?
>
> if we turn off reusing, all will actualy calculate ~count samples for each
> pixel - we will get normal of each 50..200 ray's that create radiosity in
> this point - and we can apply specular to them.
>
> this ofcourse will slow down a lot calculations - so this should be turned
> on/off by some option...
>
> --


The problem with this implementation is that radiosity would actually
need a 360 degree view of the entire scene for each pixel.
On the first pass, every pixel would know direct illumination from glowing
objects surrounding it. But how about reflection of light? For the entire 360
degree view, you'd need to know radiosity for every pixel again. On
trace_level 2, the results sky-rocket already, when moving on to
5 or higher, the image will never complete.

To simplify this, POV-Ray shoots several rays and averages results.
But I don't have that much insight into the actual code and how
it works, so I'm just guessing here. Always remember that we're talking
about a raytracer here, and getting exact results requires a lot of rays
to be shot. And, yet further, its not that easily done, cause its a
raytracer, not a scanline-render or such. The computations get
pretty large just for the scene itself, introducing radiosity was probably
no easy task...
But if you're up to it, why don't you wait for the source when 3.5 is ready
and have a go for it? ;-) (just teasing you)

Perhaps someone from the team wants to add something, or does anyone
have the thread I mentioned earlier at hand?


--
Tim Nikias
Homepage: http://www.digitaltwilight.de/no_lights/index.html
Email: Tim### [at] gmxde

Post a reply to this message