Page 1 of 1

Refraction blurring?

Posted: Tue Sep 12, 2006 2:30 pm
by phile_forum
I'm trying to make a frosted-glass room divider using a semi-transparent plane and applying 10% refraction blurring but I'm getting poor results from Kray because the refraction blurring seems to generate huge amounts of noise resulting in a very grainy image.

I've fiddled with the Kray settings and the LW texture settings for a couple of days without eliminating the noise. I've tried photon maps, light maps, high FG settings, low FG settings, lots of photons, very few photons, etc etc etc. I've also used a thin cuboid instead of a plane. No luck.

Does anyone have any experience using refraction blurring with Kray? Is it even supported properly at the moment (1.612)?

Cheers,

Phil

Posted: Tue Sep 12, 2006 4:11 pm
by jure
Hey Phile!

Can you post an example?

I've made couple of glass doors in this scene with refraction blurring:
Image and there isn't much noise.
But when there is a light source behind glass you'll get bright noise which will go away with using AA.

Posted: Tue Sep 12, 2006 5:01 pm
by phile_forum
Here's a screenie of the scene, with a bit rendered to illustrate the problem (no AA). I'm using lightmapping, the background is white and the model is open behind the glass, so there is effectively a light source behind the glass I guess. It'll take some pretty fearsome AA to smooth out that noise though!

GI Res 1m. 500000 photons received, N=500.
FG threshold 0.0, Min rays 2000, max 0.
No irradiance cacheing.

Boosting photons, N, or FG min rays doesn't seem to help.

Posted: Tue Sep 12, 2006 5:13 pm
by jure
Hmm yes this background light is problematic. In new version you'll be able to smooth it with fast AA but right now I can't realy say what would help.

Posted: Tue Sep 12, 2006 7:22 pm
by silverlw
Just a thought.. Have you tried Photonmapping instead and still have same artifacts? If that doesnt help either 1.7 is soon to be released wich is better at everything.

Posted: Tue Sep 12, 2006 7:29 pm
by jure
Photon mapping won't help because this is related to importance sampling which is apperantly still a bit buggy. And yes next version will fix it.

Progress!

Posted: Tue Sep 12, 2006 7:37 pm
by phile_forum
Yes, I did try photon mapping and, as Jure says, it didn't help.

However, Jure's remark about light sources behind the glass being problematic gave me a new angle of attack. I built a wall behind the glass, painted it white and bounced a light off it. The result isn't perfect and it still needs some tuning, but it is a big improvement. Here's another pic. Most of it is rendered with FG rays at 1000 and no AA, but the bit in the white box is rendered FG 2000 and AA fullscreen grid 3 cone 1.5 radius.

Posted: Tue Sep 12, 2006 7:54 pm
by jure
hmm much better, yes. Here's one more idea: try using limitdr 1; tailer command - maybe that will get rid of noise. http://www.kraytracing.com/manual/kray_ ... hp#limitdr

Posted: Tue Sep 12, 2006 7:56 pm
by jure
btw. AA radius 1.5 will blur alot your edges. 0.5 is optimal. But if it works good on your glass than go ahead with it.

Posted: Tue Sep 12, 2006 10:42 pm
by phile_forum
jure wrote:btw. AA radius 1.5 will blur alot your edges. 0.5 is optimal. But if it works good on your glass than go ahead with it.
True, but it looks like I'll need stronger AA here and I usually sharpen my pics in the end anyway.

Thanks for the tailer commands suggestion. I'll give it a try.

Phil

Posted: Wed Sep 13, 2006 5:40 am
by silverlw