At 02:29 PM 10/31/2005 +0000, you wrote:
Thanks for that. I see your point about the HDR, as 256 discrete steps
covering a large displacement can't be solved through interpolation.
I'll note an optimisation I don't know if you're aware of or not. In the
given example the greyscale image ranges from 0 to something like 128. In
your bump displacement you set bump amount to -0.24. Now because the image
intensity is only halfway of the maximum range (all BMP's range 0 to 1 for
0 to 255 scale, the actual range of displament is 0 to -0.12. If you load
the bmp into PS and histogram it to the full range of 0 to 255, then set the
RS bump to -0.12, you get the same level of displacement but rendered much
faster (alternatively you can load the texture and apply a curve in RS but
the PS method has less overhead). That's because RS has to look for pixel
displacements up to the maximum, whether you use that peak level or not.
Setting a lower peak and ensuring your bumpmap covers the full spectrum
makes a big difference.
David Coombes
[EMAIL PROTECTED]
You are right about all that, it can be optimized but I'm too lazy ;)
One thing to keep in mind when working with HDR is its unclear output
range. Ideally, the middle of the range should be zero displacement. I'm
just substracting a constant now, by trial & error... it can help to reduce
the total required amount of displacement.
All in all, I don't know if they optimized displacement in SP2 but I have
the impression it's much faster now than it used to be. Yesterday I was
working on a new big scene with 2048*2048 mappings and this rendered quite
easily in <10 minutes even with high quality render settings! :)
It starts to get worse if the resolution of the mesh is increased too much.
But if it's too low there's a bigger chance of artifacts.
The main problem I had was artifacts, like here:
http://www.xs4all.nl/~ath8n0r/div/big03l_artifacts.jpg
I think they can pop up when you get too close... here's another render
where the camera is higher, that's almost free of artifacts:
http://www.xs4all.nl/~ath8n0r/div/big02.jpg
If anyone can find a way to get rid of this phenomenon I'd be happy to hear it!
Another issue: I'm still not sure if a displaced surface has the correct
slope angle in VSL. This could be a problem with slope-sensitive materials.
Something to check out tonight... I'm also curious if 4096*4096 is still
manageable ;)
happy rendering,
Mark H
