Noise distortion / bulge map

Come here for help & support.
User avatar
lycium
Posts: 1216
Joined: Wed Sep 12, 2007 7:46 am
Location: Leipzig, Germany
Contact:

Re: Noise distortion / bulge map

Post by lycium » Sat Aug 14, 2010 6:55 pm

I'm currently looking at this.

User avatar
dcm
Posts: 663
Joined: Sun Jan 03, 2010 1:55 am

Re: Noise distortion / bulge map

Post by dcm » Wed Aug 18, 2010 7:54 pm

did you found something ? :))

User avatar
dcm
Posts: 663
Joined: Sun Jan 03, 2010 1:55 am

Re: Noise distortion / bulge map

Post by dcm » Sun Dec 19, 2010 6:44 am

bump :))

User avatar
PureSpider
Posts: 1459
Joined: Tue Apr 08, 2008 9:37 am
Location: Karlsruhe, BW, Germany
Contact:

Re: Noise distortion / bulge map

Post by PureSpider » Sun Dec 19, 2010 11:54 pm

OnoSendai wrote:Sorry, haven't looked at the problem yet. I'll look at it tomorrow!
Cheers!
tomorrow-ish :lol:

Stromberg
Posts: 693
Joined: Mon Mar 02, 2009 5:01 am
Location: Norway
Contact:

Re: Noise distortion / bulge map

Post by Stromberg » Fri Dec 31, 2010 7:35 am

Would be nice to hear if there are some updates on this? :)

User avatar
fused
Developer
Posts: 3648
Joined: Fri Sep 22, 2006 7:19 am
Location: Berlin, Germany
3D Software: Cinema 4D

Re: Noise distortion / bulge map

Post by fused » Fri Dec 31, 2010 11:00 am

It was looked at. :)
Not sure if it was also fixed.

User avatar
lycium
Posts: 1216
Joined: Wed Sep 12, 2007 7:46 am
Location: Leipzig, Germany
Contact:

Re: Noise distortion / bulge map

Post by lycium » Mon Jan 03, 2011 8:50 am

Indeed, and sorry for the lack of updates in this thread. A lot of other tasks (improving subdiv, GPU, and other things for the next release) took priority.

Unfortunately it's still hard to tell if an 8 bit image should be enough, judging from the original post comparison images it might be, but maybe some extra smoothing is needed... so more tests have to be done.

We'll all be back at work soon, and this will get its due attention!

User avatar
OnoSendai
Developer
Posts: 6244
Joined: Sat May 20, 2006 6:16 pm
Location: Wellington, NZ
Contact:

Re: Noise distortion / bulge map

Post by OnoSendai » Wed Mar 02, 2011 10:41 am

Hi all,
I looked into this issue a bit.
The fundamental problem is the small bit-depth of the texture, combined with the slowly changing values, leading to quantisation and hence stepping.

I have messed with curves of the bulge map a bit to demonstrate the problem.
As you can see there are basically 'plateaus' in the image with the same pixel values, which means that the shading normal will be unchanged on the plateaus, leading to stepping artifacts.

The solutions are as follows:
* Use a higher bit-depth texture, preferably floating point, such as a 16 bit EXR.
* Use ISL, maybe a perlin noise shader.
* Using a smaller texture and a wider dynamic range so that there are no plateaus should work as well.

Cheers,
nick
Attachments
bump_steps.png
bump_steps.png (6.76 KiB) Viewed 4260 times

Stromberg
Posts: 693
Joined: Mon Mar 02, 2009 5:01 am
Location: Norway
Contact:

Re: Noise distortion / bulge map

Post by Stromberg » Wed Mar 02, 2011 6:51 pm

OnoSendai: Well i think most of us know thats the cause, but dcm said he has used 8 bit images for bump in other renderengines without problems, because they do some internal filtering to smooth it out.

User avatar
zeitmeister
2nd Place 100
Posts: 2010
Joined: Tue Apr 22, 2008 4:11 am
Location: Limburg/Lahn, Germany
Contact:

Noise distortion / bulge map

Post by zeitmeister » Thu Mar 03, 2011 4:22 am

Yes, they do.
I don't know if other unbiased engines do have that problem... but ordinary raytracers do an internal filtering.
Cheers, David



DAVIDGUDELIUS // 3D.PORTFOLIO
·
Indigo 4.4.15 | Indigo for C4D 4.4.13.1 | C4D R23 | Mac OS X 10.13.6 | Windows 10 Professional x64

User avatar
OnoSendai
Developer
Posts: 6244
Joined: Sat May 20, 2006 6:16 pm
Location: Wellington, NZ
Contact:

Re: Noise distortion / bulge map

Post by OnoSendai » Thu Mar 03, 2011 7:42 am

Indigo uses bilinear filtering, but it still doesn't help in this case.

StompinTom
Indigo 100
Posts: 1828
Joined: Mon Sep 04, 2006 3:33 pm

Re: Noise distortion / bulge map

Post by StompinTom » Thu Mar 03, 2011 7:49 am

Does Indigo use 16 bits internally for bump maps? That, plus filtering on 8 bit maps, could maybe help some of the problem...

User avatar
OnoSendai
Developer
Posts: 6244
Joined: Sat May 20, 2006 6:16 pm
Location: Wellington, NZ
Contact:

Re: Noise distortion / bulge map

Post by OnoSendai » Thu Mar 03, 2011 7:52 am

Hi Tom,
Indigo just keeps whatever format the bump was loaded in. So an 8-bit bump map stays 8-bit. It could be a solution to convert to 16-bit internally and do a bit of filtering as you said.

Silmä
2nd Place Winner
Posts: 147
Joined: Fri Jun 26, 2009 7:46 am

Re: Noise distortion / bulge map

Post by Silmä » Thu Mar 03, 2011 8:54 am

There sure seems to be something really fishy going on.

These are two (rather extreme) closeups of same spot of the same test scene with the same map. The only difference (that I know of) is that the first map is 32bit exr and the second 8 bit png (reduced from the same image in photoshop).

The 8bit one completely lacks the shadow, and in the 32 bit one there's some strange patterns (could be from photoshop too, I don't know). It seems Indigo handles 32bit bump maps in totally different way than 8bit ones.

None the less, it works pretty well as abstract art!

Still using Indigo 2.4.9 though, I don't know if there's been any major improvements since that.
Attachments
test1.jpg
32 bit bump map
test2.jpg
8 bit bump map

FakeShamus
Posts: 512
Joined: Wed May 02, 2007 11:34 am

Re: Noise distortion / bulge map

Post by FakeShamus » Thu Mar 03, 2011 9:28 am

I've had a feeling there was something screwy about this for a while. noticed it a while back with my mountains scenes, where even with 32 bit maps you get stair-stepping (see the flat areas of the valley in this image).
I think this is kind of a bad problem, especially if you're rendering large or close up. isn't there a way to better translate the image data into smoother surfaces? somehow averaging the surface normals based on the differences in pixel values? (I realize in reality this is probably much more complicated than I am describing).
Attachments
5alpine3.jpg

Post Reply
42 posts

Who is online

Users browsing this forum: No registered users and 45 guests