View Full Version : GI question

03 March 2002, 08:33 AM
Ok I am sure this question gets asked a lot and and I know its pretty vague and beginner. Im sorry : \ When I searched I found some answers but mostly just got more confused by what I read. What exatly is GI?
Is global illumination just a diffuse light model? a rendering process which figures illumination by diffuse light? Is it a form of raytracing or is it hybrid scanline/raytrace or is it its own beast alltogether? And finaily is it the same as radiosity or is radiosity made up of GI, caustics and reflections? I am so confused : \

Thanks in advance.


03 March 2002, 01:37 AM
GI is just a method of calculating bounced diffuse light (as in real life). It uses rays & samples, so is a form of raytracing. it is backward raytracing (the ray samples a point, then works its way backwards through the scene bouncing off surfaces until it either finds a light source or runs out of bounces)

It is not the same as radiosity, although GI is often mistakenly called radiosity. Radiosity is another method of acheiving the same goal, but calculates light as if it was heat 'radiating' throughout a scene (hence the name 'radiosity')

Caustics is normally generated in a different pass altogether (using 'photon tracing') but some software now has the option of doing 'GI Caustics'

03 March 2002, 02:08 AM
Ahhhh Ok I see : D So caustics are possible using the model of radiosity or GI. I take it that caustics will be the same effect produced with a different algorithm depending on which renderning method you choose?

Soo... here is a question that is over my head : P Is it safe to say that the underlying computations in GI treat light as a particle whereas Radiosity treats it as a wave?

Thanks for the reply!


03 March 2002, 02:58 AM
Ahh grasshopper, but isnt a particle a wave & a wave a particle? Or is that a particle waving at a caustic as it radiates around a photon? ;)
BuggerdifIknow to be honest, I dont use any radiosity packages, but I think they are more 'mesh-dependent' & require specific triangulation to generate accurate information. so I would gather from that that radiosity treats light as if it 'flows' over a surface & the attenuation is calculated through geometry density. Wheras GI is more concerned with surface area that rays bounce between. & attenuation is calculated through numbers of bounces/material brightness.

Both of them are approximentations of reality though, so I dont know whether it gets down to the nitty gritty quantum physics level... :D

:Caustics - yep photon tracing is photon tracing - its done in a separate pass (usually) from the diffuse light so should be the same from a radiosity package to a GI package (assuming they use the same photon-calculation-algorhythm (throw that phrase out at parties & watch the chicks flock in :D )

03 March 2002, 09:55 AM

Thanks for the reply man! When I am at parties I usually drop a few names like Blinn or Catmull areound the punch bowl. Chicks dig it and the dudes get all intimidated.

Then if I am really feeling slick I whip out a copy of the Renderman Companion and watch them swoon.

: D


03 March 2002, 04:08 AM
Let me set some things straight here.
GI -- Global Illumination in rendering is about estimating the rendering equation. The rendering equation describes how radiant energy is distributed in a sythetic scene in a physically plausible manner.

Hence, a true GI renderer must simulate all effects inherent with the travel of electromagentic waves in the frequency spectrum of visible light. If is doesn't, it can't be called a "GI renderer". It may e.g. only cover part of those phenomena associated with light. In that respect, even 3D Studio 1.0's renderer somehow was a "GI-somehwat renderer".

For example, a renderer that supports diffuse transport and maybe specular transport too, but only comes with physical imlausible (yet ok looking) shaders featuring BRDFs like Blinn or Phong, can not be called a GI renderer, for this last cause.
Renderers like AIR, Entropy or MRay, which allow for custom shaders, are true GI renderers as soon as you use shaders with a phyical plausible BRDF (or BSSRDF or B*whatever) with them.



03 March 2002, 02:11 PM
hey... great GI info...!!! :thumbsup:
if you have anymore questions.. we have a thread in the maya comunity talking about the same thing.. we are trying to keep all gi info on the same thread,... easier to find... pluss there is some cool extra info if you want to check it out......

hope to see your questions there.. :)




though it might be better to pass it over to this forum,...
i'll se if we can move it over...

maybe it makes more sence

CGTalk Moderation
01 January 2006, 03:00 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.