Path: utzoo!utgpu!news-server.csri.toronto.edu!cs.utexas.edu!samsung!noose.ecn.purdue.edu!iuvax!shirley
From: shirley@iuvax.cs.indiana.edu (peter shirley)
Newsgroups: comp.graphics
Subject: Re: Ray Tracing and Radiosity
Message-ID: <57765@iuvax.cs.indiana.edu>
Date: 10 Sep 90 15:39:02 GMT
References: <11930@accuvax.nwu.edu>
Organization: Indiana University, Bloomington
Lines: 62

kaufman@delta.eecs.nwu.edu (Michael L. Kaufman) writes:

>This may be a stupid question, but why can't radiosity be handled by ray
>tracers?  Also, are there any archives that contain code/papers on radiosity
>that I can learn from?
>Michael

Ray Tracers can indeed do radiosity.  Check out the Sillion and Puech paper
or the Wallace et al. paper in Siggraph '89.  Also the Airey et al. paper
in proceedings of the symposium of Interactive 3D graphics (Computer
Graphics 24 (2)), and my Graphics Interface 90 paper.  Also Heckbert's
Siggraph 90 paper.

If all you want is a brute force radiosity code (and this code works pretty
well), then start with N polygons.  Each polygon will have reflectivity Ri
and will emit power Ei.   Assume you want to send R rays on the first pass.
We will now do what amounts to physical simulation:

T = 0              // T is total emitted power
For i = 1 to N
	Ui = Ei    // Ui is unsent power
        Pi = Ei    // Pi is total power estimate
        T = T + Ei
        
dP = T / R         // dP is the approximate power carried by each ray
For b = 1 to NumReflections
   For i = 1 to N
	r = int(Ui / dP)  // patch i will send power Ui in r rays
        dPi = Ui / r
        Ui = 0
        for j = 1 to r
		choose random direction with cosine weighting and
 		send ray with until it hits polygon p (see Ward et al.
		Siggraph 88 paper equation 2 p 87)

		Up = Up + Rp * dPi
		Pp = Pp + Rp * dPi
	

// Once this is done, we will have the power Pi coming from each surface,
// Now we should convert to radiance (see my GI 90 paper or Kajiya's
// Course notes in Siggraph 90 Advanced RT notes)

For i = 1 to N
	Li = Pi / (pi * Ai)  // Li is radiance, pi is 3.14..., Ai is area 


This ignores interpolation to polygon vertices to give a smooth image 
(see Cohen & Greenberg siggraph 85 figure 8A).




You might also want to check out Ward et al.'s Siggraph 88 paper.  Not true
radiosity, but is ray tracing based and has some of the best features
of ray tracing methods.  If your scene is all diffuse, that method may be
the way to go.  

Enjoy,

Pete Shirley
shirley@cs.indiana.edu