Visual Chatter (Direct vs. Global Ill.)

(L) [2006/06/27] [toxie] [Visual Chatter (Direct vs. Global Ill.)] Wayback!

The BOMB:
[LINK http://www1.cs.columbia.edu/CAVE/publications/pdfs/Krishnan_EUROGRAPHICS06.pdf]

Very, very interesting paper. He extracts the direct and global illumination from real life scenes using a rather simple trick.

(L) [2006/06/27] [playmesumch00ns] [Visual Chatter (Direct vs. Global Ill.)] Wayback!

This is fascinating. I work as a shader developer in the vfx industry and having this technique avaiable when designing material solutions would be incredible.

Anyone read and understood the papers they cite on the practical implementation

(L) [2006/06/27] [beason] [Visual Chatter (Direct vs. Global Ill.)] Wayback!

They also have a siggraph paper on the same topic:

[LINK http://www1.cs.columbia.edu/CAVE/publications/pdfs/Krishnan_TOG06.pdf]

Not sure what the difference is (haven't read either).

(L) [2006/06/28] [toxie] [Visual Chatter (Direct vs. Global Ill.)] Wayback!

The tricky thing is that most of the presented/photographed stuff requires a accurate simulation of subsurface scattering and especially spectral rendering to deliver equal results (f.e. the flower that looks white under direct illumination and only "changes" color due to massive number of bounces and thus "changing" the spectrum of the incoming light).

(L) [2006/06/28] [playmesumch00ns] [Visual Chatter (Direct vs. Global Ill.)] Wayback!

You mean in simulating it in CG? There's plenty of ways of getting the same look [SMILEY :)]

So as I understand the technique, you shoot an image using a single light source that has a checkerboard gobo on it. Parts of the image where the checkerboard is blocking the light will therefore be receiving only global illumination. Then you shoot another one using a gobo that is the inverse of the first. Combining the "dark parts" of those two images gives you the global illumination only, then presumably you just subtract the global image from the combined "bright parts" image to get the direct on it's own?

Very clever [SMILEY :)]

(L) [2006/06/28] [toxie] [Visual Chatter (Direct vs. Global Ill.)] Wayback!

exactly how you said.

And for your first comment:
But faking the ssscattering and spectral rendering just destroys the whole point of being able to model "real" scenes using some photographs.
I would have never thought that some materials look the way they look in this paper (=material like organic stuff (corn, skin, etc.) under direct illumination only). So i think it helps all of us coders to make better/more realistic test scenes to test our GI-stuff.

(L) [2006/06/28] [playmesumch00ns] [Visual Chatter (Direct vs. Global Ill.)] Wayback!

I suppose if I was concerned with absolute physical accuracy that would be true [SMILEY :)]

I'm more concerned with just getting the look of a certain material right. Being able to seperate the direct and indirect components of reference photographs would be an invaluable aid in trying to figure out how best to approximate that material.

I think I'm going to try it out, and if I can get a decent result using blue peter-style equipment I may just try and do it for a real project...

(L) [2006/06/28] [playmesumch00ns] [Visual Chatter (Direct vs. Global Ill.)] Wayback!

You mean in simulating it in CG? There's plenty of ways of getting the same look [SMILEY Smile]

So as I understand the technique, you shoot an image using a single light source that has a checkerboard gobo on it. Parts of the image where the checkerboard is blocking the light will therefore be receiving only global illumination. Then you shoot another one using a gobo that is the inverse of the first. Combining the "dark parts" of those two images gives you the global illumination only, then presumably you just subtract the global image from the combined "bright parts" image to get the direct on it's own?

Very clever [SMILEY Smile]

(L) [2006/06/28] [playmesumch00ns] [Visual Chatter (Direct vs. Global Ill.)] Wayback!

I suppose if I was concerned with absolute physical accuracy that would be true [SMILEY Smile]

I'm more concerned with just getting the look of a certain material right. Being able to seperate the direct and indirect components of reference photographs would be an invaluable aid in trying to figure out how best to approximate that material.

I think I'm going to try it out, and if I can get a decent result using blue peter-style equipment I may just try and do it for a real project...

(L) [2006/06/29] [Guest] [Visual Chatter (Direct vs. Global Ill.)] Wayback!

Cool.. Looking forward to hear from you if this works sometimes (and you are allowed to post some pictures [SMILEY :)] !

Visual Chatter (Direct vs. Global Ill.) back