Aliasing effects (antialiasing)

Phil · Post by **Phil** » Tue Nov 14, 2006 9:18 am

Hello everybody,

It is relly a pleasure to work with indigo

but I have new problems.

I have some strange aliasing effects, and.. yes I know there is no antialiasing parameter in indigo.

The first picture after about 7 hours rendering.

The second after about 25 hours rendering

I do not see a quality improvement over the rendering time, so I have some questions now:

1) are the units playing a rule?
2) does it exists some parameters to improve this?

Thanks in advance for all answers or suggestions!

regards,
Phil

StompinTom · Post by **StompinTom** » Tue Nov 14, 2006 11:13 am

best way to get around any AA problem is to render twice the size and scale down afterwards

works everytime....
a bitch for rendertimes though...

tungee · Post by **tungee** » Tue Nov 14, 2006 8:42 pm

which version of indigo used for rendering?

Phil · Post by **Phil** » Tue Nov 14, 2006 8:55 pm

I use 0.6 test 6

tungee · Post by **tungee** » Tue Nov 14, 2006 9:01 pm

hmmm...

zuegs · Post by **zuegs** » Tue Nov 14, 2006 9:45 pm

I think it's tue to the normal light energie integration.

If you average a serie of energies and then clamp to [0..1]:
(0.5 + 0.5 + 4 + 5) / 4 = 2.5 -> gets clamped to 1.0

but when you first clamp energies and then average them:
(0.5 + 0.5 + 1.0 + 1.0) / 4 = 0.75

Indigo uses the first averaging because second can't be used (i think

). That means indigo clamps pixels after averaging and tonemapping what finaly produces sharp edges on high illuminated areas.

That's why StompinTom's oversize rendering with downscaling (matchs second averaging model) fixes some of this AA problems

I don't know if anyone will understand my ugly english-post

but I can't explain better

regards zuegs

tungee · Post by **tungee** » Tue Nov 14, 2006 10:51 pm

could ono fix this or is this indispensable zuegs?

Post by **OnoSendai** » Tue Nov 14, 2006 10:58 pm

Yeah, I'm pretty sure such an effect is causing the aliasing Zuegs, i.e. it's caused by the non-linear tonemapping function being applied after the sample filter patterns are summed together.

I have some ideas on how to fix this, but if anyone has any relevant papers etc.., flick them my way please

zuegs · Post by **zuegs** » Tue Nov 14, 2006 11:00 pm

I don't know if there's a fix possible

curently i can just say it's the nature of mathematical-"integration" and in this case it's just correct as it is

, even if it's not nice for the eyes

. Perhaps some special saturation-tonemapping could reduce the effect a bit... or you can perhaps use less sharp light-sources (sphere instead of plane or circle with a natural fall-off to the light-border)

tungee · Post by **tungee** » Wed Nov 15, 2006 12:09 am

he ono here are some papers for you!
http://rapidshare.com/files/3309695/tun ... A.zip.html

and some more
http://rapidshare.com/files/3310275/new_papers.zip.html

fix da bugs

Post by **OnoSendai** » Wed Nov 15, 2006 12:16 am

Thanks Tungee.. but ... zipped on rapidshare?

can't u just post links ?

tungee · Post by **tungee** » Wed Nov 15, 2006 12:20 am

i thought you dont need to dowload these seperately and get disoriented by the links

Phil · Post by **Phil** » Wed Nov 15, 2006 12:36 am

Thanks all for your reply,

I will try the method of StompinTom.

An another which cames on mind is that the radius here is very small.
The width is about 50 mm and the radius is about 0,2mm.
Perhaps is there also some minimal resolution to apply for rendering this correctly, I don´t know.

Regards, Phil