I'm curious what people are using for omni-directional shadow maps these days.
I think that dual paraboloid and cube map shadows are two typical choices. I have used the latter in the past. While looking for alternatives that might perform faster with comparable quality I came across Dual Sphere-Unfolding Shadow Maps, which are rendered in a single pass. Has anyone tried it? Is there a more detailed explanation of the method around? Any other methods that are worth looking into?
3
Dual Sphere-Unfolding/Omni-directional shadow maps
Started by B_old, Jul 19 2012 07:07 AM
7 replies to this topic
Ad:
#2 Members - Reputation: 281
Posted 19 July 2012 - 05:19 PM
I tried this method in a recent project to fill a scene with 100 dynamic lights casting shadows. The theory is appealing and the results are not so bad...
BUT you'll want to have a very fine geometry in order to obtain accurate shadows: the projection used to store the whole shadow sphere in a single texture will highly bend the geometric data, thus leading to frequent artifacts during the depth comparison.
(Though I had to rush that project so I might have missed something)
Still, it was tremendously fast and produced cool shadows... If you don't stare at them too much
BUT you'll want to have a very fine geometry in order to obtain accurate shadows: the projection used to store the whole shadow sphere in a single texture will highly bend the geometric data, thus leading to frequent artifacts during the depth comparison.
(Though I had to rush that project so I might have missed something)
Still, it was tremendously fast and produced cool shadows... If you don't stare at them too much
Edited by PixelSmasher, 19 July 2012 - 05:21 PM.
#4 Members - Reputation: 281
Posted 24 July 2012 - 03:12 AM
Unfortunately, I am not.
This technique seems to remain unnoticed (or people consider that such artifacts are unworthy of their time).
There is this video, made by the author of Dual-Sphere Unfolding (It was my only help when implementing my demo), but nothing from the community.
If you ever find something, I would be greatly interested.
This technique seems to remain unnoticed (or people consider that such artifacts are unworthy of their time
).There is this video, made by the author of Dual-Sphere Unfolding (It was my only help when implementing my demo), but nothing from the community.
If you ever find something, I would be greatly interested.
Edited by PixelSmasher, 25 July 2012 - 06:23 AM.
#5 Members - Reputation: 712
Posted 25 July 2012 - 09:43 AM
It's just simply sphere mapping. The only difference is, that they use 2 sphere maps and let the vertex shader decide on which one the vertex gets projected, based on whether the y axis is positive or negative. They say that it would be a one pass method, but I really doubt it. I don't think that a triangle that has vertices on both sides of the xz-plane wouldn't cause artifacts. You'd probably still need a geometry shader or 2 passes to render these shadow maps without artifacts. I think it's worse than dual paraboloid shadow mapping, but I'll give it a try.
#6 Members - Reputation: 373
Posted 26 July 2012 - 04:47 AM
It would be nice to see your results in case you find the time and implement it.It's just simply sphere mapping. The only difference is, that they use 2 sphere maps and let the vertex shader decide on which one the vertex gets projected, based on whether the y axis is positive or negative. They say that it would be a one pass method, but I really doubt it. I don't think that a triangle that has vertices on both sides of the xz-plane wouldn't cause artifacts. You'd probably still need a geometry shader or 2 passes to render these shadow maps without artifacts. I think it's worse than dual paraboloid shadow mapping, but I'll give it a try.
#7 Members - Reputation: 523
Posted 26 July 2012 - 08:45 PM
I would also try single pass cubemapping using the geometry shader: http://devmaster.net/forums/topic/9367-render-to-cubemap-in-a-single-pass-really-useful/
The least amount of geometry warping that I know of, and still very efficient
The least amount of geometry warping that I know of, and still very efficient
#8 Members - Reputation: 373
Posted 30 July 2012 - 02:52 AM
Last time I tried it, using the geometry shader for single pass cubemapping was slower on my machine than doing a pass for every cube face. Have you seen more encouraging results?I would also try single pass cubemapping using the geometry shader: http://devmaster.net...-really-useful/
The least amount of geometry warping that I know of, and still very efficient
