PDA

View Full Version : Shadows: a question


Stoehr
06-25-2003, 10:41 PM
In the attached image you can see a spot of light inside the nostril. Technically, with my shadow casting light coming from the left side, there should be no interior lighting of the nose.

1. I'm using shadow maps: Shadow Res. 3000 Shadow fuzziness 3.0
2. There are three lights total:
a. Key - shadow only, at 100%
b. Fill - no shad., 50%
c. Rim - no shad., 100%
3. No ambient intensity
4. No radiosity
5. Shading noise reduction is "ON"

I remember a little about this topic mention months ago, and I think it has to do with zooming in too much on the polygons.

Can someone confirm this issue? Or shed some light? (no pun intended)

Thanks.

E_Moelzer
06-25-2003, 11:00 PM
I think this is because of the problems Shadowmaps ahve with concave objects. Solution: Dont use shadowmaps! They suck anyways.
CU
Elmar

MorBioS
06-25-2003, 11:02 PM
sometimes I have it as well.

It's a lwbug. I see that since lw 5.6.

Try use a raytrace light or area light

Mattoo
06-26-2003, 12:15 AM
It might also be that you've got a 100% rim light that isn't casting a shadow. Hence you've got polys on the front side of the model that are recieving light through the head.

You're example could be a poly that is catching that light?

Btw, I use shadow maps all the time, especially the spinning variety - very useful.

Stoehr
06-26-2003, 12:16 AM
Well, after more testing I found the raytracing of shadows did not alleviate my problem. Somehow my fill and rim lights were lighting "through" my object. Causing tiny highlights on edges parallel to the lights. I know this has happened with object with holes. But I closed any holes (ie. the mouth, noes, ear, eyes), and I still have lit nostrils.

To fix this issue I could use a second shadow light in the Fill light. However, this solution is only good for this particular portrait. In the end, an area light will be the most accurate, but the most time consuming. So I'll end up spinning my lights.

Werner
06-26-2003, 03:00 PM
place a point light in each of the nostrils and change their intensity to - 100% . Let me know if it works.

CourtJester
06-26-2003, 03:33 PM
Originally posted by MorBioS
sometimes I have it as well.

It's a lwbug. I see that since lw 5.6.

Try use a raytrace light or area light

Raytraced lights have bugs of their own, in particular random speckles of light in the shadows, especially with subpatched objects. I myself have been seeing that since 5.0 (I still have the test image I submitted to Alan Chan back in 1996).

Anybody suppose that it's due to distant lights, and isn't an issue with RT spots? I must test.

Sterk
06-26-2003, 04:41 PM
I have similar problem every time when modelling human head and till now I found two solution for this
1 Use negative point lights as was mentioned
2 use weight maps on gradient which assigned to diffuse channel of head surface
ussually the second option gives better result:wavey:

WChestnutt
06-26-2003, 05:04 PM
Stoehr, have you turned on double sided polys?

i have found that having double sided polys off can cause light to shine through. (i might be wrong though...)

Zarathustra
06-26-2003, 06:17 PM
I think unknown limit hit it on the head. I've experienced that same issue before and making it double sided worked.

Stoehr
06-26-2003, 06:43 PM
Originally posted by unknown limit
Stoehr, have you turned on double sided polys?

i have found that having double sided polys off can cause light to shine through. (i might be wrong though...)

Alright, double sided polys works, however, is this method valid in production? Wouldn't the render times increase with all things that calculate based on the surface normals?

Stoehr
06-26-2003, 06:45 PM
Originally posted by Sterk
I have similar problem every time when modelling human head and till now I found two solution for this
1 Use negative point lights as was mentioned
2 use weight maps on gradient which assigned to diffuse channel of head surface
ussually the second option gives better result:wavey:

The weight maps idea sounds good. Then a light rig doesn't need to be made and managed.

WChestnutt
06-26-2003, 08:13 PM
i would think that double sided would be the norm, i haven't heard of any other way.

Will

CGTalk Moderation
01-15-2006, 01:00 PM
This thread has been automatically closed as it remained inactive for 12 months. If you wish to continue the discussion, please create a new thread in the appropriate forum.