Human Head (banding?).

[Mendel]

I ran into an image quality anomaly. What do you guys think is happening here?



Click to open image, look at his left eye. (right from your pov)

(yes, I have moved the ambient slider)

edit: Just to note. No, it's not a problem, I'm not bitching, just curious..

edit2: mods may feel free to move this to the beginner's question forum. This might have been more appropriate there!

 

[digitalwanderer]

It almost looks like you can see his eye/eyeball faintly through the eyelid, is he looking hard left (right from the viewers perspective) underneath that closed eye?

 

[Mintmaster]

Ahh yes, LCD owners with their crappy black levels and messed up gamma curves.

If you had a real display, you wouldn't see anything wrong.

 

[Andrew Lauritzen]

If you had a real display, you wouldn't see anything wrong.

Looks fine on two different LCDs here, but actually that image is a bad example. There is indeed some rather severe banding and artifacts from certain angles in that demo, and that was on a CRT I didn't spend the time to figure out what was going wrong though as the artifacts are pretty rare and probably easy to make more robust if necessary.

 

[Jawed]

Bands on my gamma 2.2 calibrated CRT, too.

Jawed

 

[nAo]

Screwed tone mapping or screwed HDR image with a very very very small dynamic range.

 

[Mintmaster]

Bands on my gamma 2.2 calibrated CRT, too.

In that screenshot? You sure you're seeing bands?

If you run the demo, you can see that the base texture is a different colour on part of the left eye even when well illuminated. My guess is it looks out of place when it isn't as dark as it should be, and by jacking the brightness of my monitor I can see how it would be mistaken for banding.

Andy, I haven't seen the bands that you're talking about, but I haven't been analyzing the demo either.

 

[Andrew Lauritzen]

If you run the demo, you can see that the base texture is a different colour on part of the left eye even when well illuminated. My guess is it looks out of place when it isn't as dark as it should be, and by jacking the brightness of my monitor I can see how it would be mistaken for banding.

Yeah looking at the RGBs in paint shop pro there are no real discontinuities (although there are discontinuities in the derivatives, but that's to be expected with such a texture map), so there shouldn't really be "banding".

Andy, I haven't seen the bands that you're talking about, but I haven't been analyzing the demo either.

I haven't analyzed it much either to be honest - it was just at work when I was showing it to a few people that I happened across an angle that both some really ugly bands and some really bad shadow-map-like aliasing of some function related to the shading (could have actually been the same artifact come to think of it... I didn't look into it).

[Edit] Here is an example of the projection-like artifact that I saw, although it was much more obvious when we saw it initially. Also the banding isn't really present in this image, although again it was when I initially noticed the artifact(s).

 

[Mintmaster]

Yeah, you're right. It's just the same shadow map artifacts you always see on angled surfaces. The big squares are from the polygons of the head, and maybe that affects gradients or something else. The small streaks are from the texels in the shadow map.

As you're probably well aware of, an edge far from vertical or horizontal give you a comb of depth values that sort of scrape the low angled surface during projection.

EDIT: Never mind, it's something else. Maybe iterator or shadow map precision.

 

[Jawed]

In that screenshot? You sure you're seeing bands?

I'm seeing bands in the screenshot posted. Now why they are there to be seen is a different question - one that I've got zero opinion on, plus the fact I can't run the demo.

I was merely reporting my CRT datapoint.

Jawed

 

[Mendel]

If you can run the demo just zoom closer to that eye and move the ambient slider back and forth. Should be pretty obvious what I mean. I tried capturing a video about it but gave up on trying to make the file size small enough for posting.

 

[Mintmaster]

Mendel, those aren't bands on the eye. It's just the way the texture art looks like with the left eyelid having a rather whitish spot. If you want to see for yourself, you can open the file "textures\AdrianAlbedo.tga" in the demo's install folder. It's a huge 65MB uncompressed texture.

(heh, too lazy to change the name away from "Adrian" )

I'm seeing bands in the screenshot posted.

In the same location that Mendel is talking about or elsewhere?

If you have gamma 2.2, I don't know how they can be visible enough to call them bands. The RGB values are around 50, so that equates to under 3% luminance, and if you use an image editing program, you can see that there are plenty of colours between that and the surrounding darker areas.

 

[Jawed]

If you have gamma 2.2, I don't know how they can be visible enough to call them bands.

The head's left eyelid looks like it has "banding". As I said earlier, I don't know why, I'm just reporting an apparent artefact that's visible here. I can't engage in any image anlysis right now because I'm re-building my system after my system disk kicked the /dev/null bucket.

Jawed