PDA

View Full Version : Radiosity: Still Geometry-dependent?


jeremybirn
09-19-2005, 06:33 AM
I posted this on the Lighting & Rendering forum, but wanted to hear definatively about the C4D side of things.

I've been using other GI approaches for a few years, and I haven't seen what's going on with implementations of Radiosity.

Is it still geometry-dependent as it was in Lightscape, where irradiance info was stored in vertices of the geometry, and you got slower but more accurate solutions on surfaces with higher polygon counts? Back in Lightscape, the software even had to adaptively tesselate lower-poly surfaces to add vertices where there was high contrast in the irradiance data - if you use radiosity does that still happen in modern implementations?

I'm only asking about Radiosity, not other GI approaches.

-jeremy

Per-Anders
09-19-2005, 07:16 AM
Cinema only has GI inbuilt I'm afraid (Stochastic brute force, and some form of photon based). It's largely mesh resolution independent (but not mesh detail independent). I've put a post in the Lighting & Rendering forum that should give a clearer explanation.

jeremybirn
09-19-2005, 07:45 AM
Thanks for your continued help! I see many references to "Radiosity" on this forum, and am writing something that needs to be 100% accurate so I wanted to make sure I don't make any over-generalizations about anything that's labelled radiosity in any popular program. If they have taken that word out of C4D's interface then I have nothing to worry about.

-jeremy

Per-Anders
09-19-2005, 07:49 AM
yes, the word has now been removed. The interface now correctly states "Global Illumination" (though i've not checked the manual yet, though I presume it's been corrected there too).

An Erased One
09-19-2005, 10:40 PM
Isn't the GI in Cinema a form of Radiosity? It only calculates diffuse lighting, and the distribiution of samples is very similar to the "classic" approach lighscape seemed to do - "only" that not the faces, but samples on the surface are used to determine the solution.
The implementation is different, but the idea seems to be the same.

Per-Anders
09-19-2005, 11:22 PM
From an algorithm standpoint they're quite different. Radiosity is forwards casting, i.e. it goes from lightsources. GI is backwards casting, i.e. it starts out from the eye ray, this requires quite a different approach.

The way the samples are calculated is also quite different. You'll notice that the samples Cinema creates are clustered around detail, this is a very different algorithm to simply depending on how dense your mesh is to determine where samples should be.

In the end the results should ideally be very similar, they are both methods of simulating and calculating bounced or ambient light, GI has the benefit of focusing solely on what's in viewport with minimum samples outside, while radiosity inherently calculates the whole scene.

An Erased One
09-20-2005, 01:24 AM
Ah, okay, that makes sense. I never really looked at the prepass, but I always thought it was just a way to determine how to put samples on the geometry for a form of radiosity - since it was labeled "radiosity" up to ver. 9.1 ;). Thanks for the Information.

jeremybirn
09-20-2005, 03:53 PM
From an algorithm standpoint they're quite different. Radiosity is forwards casting, i.e. it goes from lightsources. GI is backwards casting, i.e. it starts out from the eye ray, this requires quite a different approach.

Although I hope you agree that Radiosity is one approach to Global Illumination.

-jeremy

Jorge Arango
09-20-2005, 05:54 PM
GI has the benefit of focusing solely on what's in viewport with minimum samples outside, while radiosity inherently calculates the whole scene.


Has any of this changed with AR 2.5, now that it's officially GI or do outside objects influence the calculations? You say "with minimum samples outside" but definitely there is noticeable influence in AR 2.0.


Jorge Arango

dann_stubbs
09-20-2005, 07:12 PM
Although I hope you agree that Radiosity is one approach to Global Illumination.

-jeremy

hey jeremy, nice to see you around this forum.

dann

Per-Anders
09-20-2005, 08:09 PM
Although I hope you agree that Radiosity is one approach to Global Illumination.

-jeremy

Yes, it is a form of Global Illumination, however strictly speaking we're running out of terms to call the overall effect :D (it gets worse when people call it ambient lighting too as there's already a bunch of ambient parameters in any 3d app).

Really either Global illumination as an algorithm versus Radiosity as an algorithm, or Global Illumination as a description for the effect of bounced light needs to be differentiated as it's come to mean two things, one of which is exclusive of certain forms of the other!

jeremybirn
09-21-2005, 07:30 AM
Yes, it is a form of Global Illumination, however strictly speaking we're running out of terms to call the overall effect :D

Global Illumination is the overall effect. Research papers and anything I've seen that's carefully edited use "Global Illumination" as the overall, blanket concept, without regard for which type of global illumination algorithm you use.

AR 2.5's GI solution is a kind of global illumination. Radiosity is a kind of global illumination algorithm. Photon Mapped GI is another kind of GI algorithm. And so on.

(I mentioned this on the other thread, but a lot of mis-use of terms got started in the 1990's because the first broadly available commercial GI solution was Radiosity, so a lot of people started over-using that term as if it were a general description of bounced light. Really radiosity is just one type of GI algorithm, the term only got overused because it came out first.)

-jeremy

CGTalk Moderation
09-21-2005, 07:30 AM
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.