BB FORUM - Powered by X M B 1.8 Partagum Final SP1

Please Enable Javascripts to view our site Properly.

Not logged in [Login - Register]

BB FORUM » 3D » Radiosty 101 (A MUST READ for anyone doing 3d) » Post Reply

Who Can Post?

All users can post new topics and all users can reply.

Username

Need to register?

Password:

Forgot password?

Subject:

(optional)

Icon:

Formatting Mode:
Normal
Advanced
Help

Message:
HTML is On
Smilies are On
BB Code is On
[img] Code is On

[quote][i]Originally posted by snuggs[/i]
[b]Radiosity 101.[/b]

First of all i wouldn't call this a tutorial because i don't like that word.
In the 3d world we tend to limit ourselves by the application we use
as apposed to the concept of what we are doing. When you think of radiosity you must get away from the 3d application you are using.
Once again i cannot stress this enough.  I have found information writen on maya and converted it to 3ds max and found great stuff on max that i used in lightwave.  Same thing with renderers, same thing with programming languages.  They all are trying to solve some real world equation.  I sometimes take code logic i've writen in C# or even COBOL (yes cobol) and convert the logic to PHP or Actionscript...it's all just math, but that's beyond the scope of this.

[b]By Definition.[/b]
The main goal of radiosity is very simple.  To provide light rays to areas
where conventional 3d lighting does not reach but would reach in the real world.  Radiosity occurs when photons strike a material and bounce off it in another direction.  In the real world, radiosity light reflections happen hundreds or thousands or millions of times until all the energy is absorbed.  A nice example of where you see radiosity is with indirect illumination bounces
you might see under a person's chin on a bright day if they are over something white.  Light bounces off the table, off the chin..into your eye.  JORDAN FOR 3!!!

This has been called "Ambient light" in artificial illumination definitions. 
(NOT TO BE CONFUSED WITH AN AMBIENT LIGHT IN A 3D APPLICATION).

Global Illumination with Ambient Occlusion using Maya 6.5 Unlimited
[img]http://www.monstaronline.com/_uploaded/ambient_occlusion.gif[/img]

This method has been faked for years but an indoor scene without radiosity is like an outdoor scene without an High Dynamic Range Image illuminating/casting reflections onto your model.  Now of course we could not calculate TRUE radiosity because computers are not fast enough to calculate each ray bouncing infinitely until it's energy wave reaches zero.  We usually are familiar with faking radiosity by using Global Illumination methods.  Caustics is another good example of refracted indirrect illumination except the light is reflected or refracted into focused high-density areas of illumination,
such as when light shines through a glass of wine and onto a tablecloth under the glass.  (Notice the coloring of the light rays going through the wine onto the tablecloth.
[img]http://www.monstaronline.com/_uploaded/wine_glass.jpg[/img]

Maya 6.5 Unlimited, water NextLimit realflow
[url=http://3dvirtualight.com/Download/wave_caustic.avi]Click Here[/url]

The one problem with GI is that it's actually quite horrible without at least 1 light (GI faking radiosity is when INDIRECT ILLUMINATION is enabled).  I mean how often do you
see an overcasted day with no sunlight?  Just doesn't look right for final renders.
But usually due to a production deadline in your pipeline you can't waste time so GI's the default. Global Illumination is an easy fake to radiosity as is a falloff Fresnell reflectance in a material fakes physically correct Raytrace reflections and saves time but with a quality tradeoff.

[b]The Problem With Conventional Lighting rigs.[/b]
Conventional lighting setups, no matter how good, do not account for one thing, diffuse interreflections. With a regular lighting setup, especially indoors, the materials and lights account for only the light arriving directly from the light sources themselves.  This does not take into account the fact that a room is not only lit from the source light objects (direct lighting), but also arrives from other surfaces (indirect lighting).
Look under a table in a well lit room.  Notice how without the light bouncing from the carpet, the legs, and walls, underneath
the table would be completely black.

[b]What's with the Star Wars?[/b]
You may hear the word photon from time to time and these are only samples of light that contain an amount of light energy used for calculation.  As a photon simulation travels (Luminance) and bounces (Illumination),  the light energy weakens just as in the real world (usually inversely in the real world...keep this in mind when using lights).
Think of it as having properties of a particle in
respect to a photon can bounce around until it loses all of its  energy. This
"energy" is projected similar to a wave when passing through a camera.  (The tecnical term is the Wave-particle Duality phenomena used in quantum mechanics and physics but i won't bore you with that)
When you use a light in your scene, one reason why your radiosity could not be working properly is because you have no lights in your scene emitting photons.  The following rollout is from Splutterfish's Brazil r/s Photon Maps rollout when you enable photometric lighting.
[img]http://www.monstaronline.com/_uploaded/photon_light_rollout.gif[/img]

[b]Color Bleeding without the bandaids.[/b]
[img]http://www.monstaronline.com/_uploaded/color_bleeding.jpg[/img]

Materials absorb photons and send them along their way in real life lighting.  A Cornell Box is a perfect setup of what our eye adjusts to without us noticing.  The colors to either wall are usually complementary colors and are placed next to a white wall (the back wall) to do color bleeding tests.  As light photons bounce around they also carry some of the diffuse color with them to the adjacent walls.  And if your corners are too dark, usually bumping up your photons, decreasing your gather radius (making what adjacent photons to include with itself smaller to "draw" light closer to it), and increasing the bounces slightly will fix this.

[b]What's with the Colorful Room?[/b]
[img]http://www.monstaronline.com/_uploaded/cornell_comparison.jpg[/img]

Well, actually it is'nt a room, it is called a "Cornell Box".  Founded by
Cornell University, the box was used to view the interaction of light between diffuse surfaces.  which a perfect one will have 49 quadrillateral faces per side but MOST IMPORTANTLY, is physically correct in scale.  The physical cornell box was used to depict how light traveled in small, enclosed spaces.  I usually use a Cornell Box when i am done modeling and i am doing light tests.  Most people go with global illumination using a skylight which is kinda sorta correct for outdoor drab scenes but I usually prefer Ambient Occlusion over the typical Skylight, and that is only for outdoor scenes.  That is why alot of indoor scenes with this lighting setup don't look right because light isn't traveling correctly through the scene, nor are there any "skylights" in an indoor scene.  most light coming through a room in daytime is through a nearby window and bounces around in the room until it loses energy.  Even at night, at lowest the light
comes from moonlight outside or an incandecent light inside but Radiosity is still present, NOT SKYLIGHTS.  Skylights do not provide a good angle of incidence either.

[b][url=http://www.cgtalk.com/showthread.php?t=233460&page=1&pp=15]CLICK HERE[/url][/b] FOR AN AMAZING LOOK AT RADIOSITY AT DAYTIME AND NIGHT TIME WITH A FLASHLIGHT SETUP...
YOU WONT BELIEVE IT'S NOT REAL!

The Angle of incidence (the angle at which the light photon will collide/leave with the surface.  BRDF (Bidirectional Reflectance Distribution function). 
These angles are usually the same unless light is being bounced around inside the surface and leaves at a different and/or multiple angles. You can see this happening with say for instance a bunch of grape, or human skin (hold your hand up to a light and look at it glow) 
BSSRDF (Bidirectional Sub-surface Scattering Reflectance Distribution Function).
You can learn more about these functions by reading this Siggraph paper on the topic
[url=http://graphics.stanford.edu/papers/bssrdf/bssrdf.pdf]Click Here for University of Stanford Siggraph Proposal[/url].

Radiosity reflections can continue infinitely until all photonic energy has been spent.  This is what happens in the
real world and is reproduced using physically correct lighting in your 3d application.
PHYSICALLY CORRECT LIGHTING MEANS YOUR SCENE MUST BE DONE TO PROPER SCALE OR
THE WHOLE THING WILL NOT WORK.  Not to mention if it's scaled too large you will wait forever for a perfect render and will not happen.  Imagine if you were a photon bouncing around in a gymnasium compared to a photon bouncing around in a box of matches.  Which one will lose energy the easiest.  BUT, if you render a gymnasium the size of a box of matches or vice versa, not only are your calculations incorrect, you will be able to tell with your renders, especially the ones with semi Translucent Sub-Surface scattering materials put on them such as skin, wax, marble, or plastic.   So basically, [b]the most important thing to remember is if your wall is supposed to be 6 feet tall, make it 6 feet tall not 60, you will live to regret it if you don't[/b].

If you would like more background information on the cornell box,
[url=http://www.graphics.cornell.edu/online/box/]Click Here[/url].

Richard Rossman the creator of [b][i]The Plumber[/i][/b], one of my favorite CG Artists has a great Cornell Box explaination
[url=http://www.richardrosenman.com/cornell.htm]HERE[/url].  
You can also check out an interview with him at CG Talk [url=http://www.cgnetworks.com/story_custom.php?story_id=2858&page=]HERE[/url]

Cornell Box Scene - http://www.vrayrender.com/stuff/PMapTutorial/cornellFinal.zip

[b]Selecting the Perfect Renderer.[/b]
The downside of radiosity is usually you must calculate a "Radiance Map" of your scene before you can render the scene.  The great part is once you have calculated the photons bouncing, usually you can use this same calculation even when switching camera angles.  Usually this is computed using the Monte Carlo / Quasi-Monte Carlo methods no matter the 3d application which calculates the sampling of each pixel as the light travels through the scene and finally into the camera. Usually undersampling this and progressively lowering the sample values (actually raising them because they should
never break the range between -5,0 and 0,1 or your render times will boost drastically.
Vray Quasi-Monte Carlo Rollout
[img]http://www.monstaronline.com/_uploaded/vray_qmc.gif[img]
Brazil Quasi-Monte Carlo Rollout (Notice the bounces talked about earlier)
[img]http://www.monstaronline.com/_uploaded/brazil_qmc.gif[img]
 
[i]Scanline.[/i]
Actually, there are none.  Say you have only the default scanline renderer
in 3ds max.  If you go under the Raytracer tab...
[img]http://www.monstaronline.com/_uploaded/scanline_raytrace.gif[/img]
Here is the rollout that you need for enabling raytracing.
under Advanced Lighting Select the Radiosity plugin.  Radiosity settings here
are pretty self explainatory if you understand what is going on behind the scenes. Click the Start button and go get a cup of coffee, or answer some emails and the solution should be done by then.
If your scene is scaled proportionately it shouldn't take that long to create a solution.

If the image doesn't come out the way you'd expect then try messing with
the Exposure control (Rendering->Environment or Press 8).
[img]http://www.monstaronline.com/_uploaded/environment_rollout.gif[/img]
I've found Automatic exposure control to be the easiest to use, but once
you become familiar with the other methods, you can produce some pretty
interesting renders.

[i]Mental Ray (works the same in 3ds Max and Maya).[/i]
Mental Ray calculates its indirect illumination in terms of units called
photons (just like every other method of radiosity).  Mental Ray uses two types of photons; global illumination photons and caustics photonÃ¢â‚¬â€?and requires photon shaders to be used in the scene. Simple ray trace materials will work fine.
Maya 6.5 Unlimited
[img]http://www.monstaronline.com/_uploaded/cornell_s.jpg[/img]

[i]Vray r/s.[/i]
Uses Radiance Maps which areas appear to be darker due to some miscalculations
when doing Radiance/Irradiance Maps.  This method may be the fastest but not the best.
[img]http://www.monstaronline.com/_uploaded/vray_irradiance_map.jpg[/img]

[i]Brazil r/s.[/i]
Brazil is by far my favorite renderer but it is a pain when working with indoor scenes.  Since Splutterfish's Brazil in my opinion comes up with better Radiosity computations, the downside is brazil does take a bit longer for renders to come out.

There are at least 3 mandatory steps to allow photon mapping to work with brazil.  First you must enable photons to be emitted from your light source.  This usually is as simple as enabling the light to be photometric
and enabling the photons checkbox from the light's properties. THE CALCULATION WILL NOT WORK WITHOUT THIS DONE!

Secondly, In brazil it is best practice to use the Brazil Utility material because you can locally and manually tweak the global illumination absorbsion of the material from within the Material Editor.
[img]http://www.monstaronline.com/_uploaded/brazil_utility_material.jpg[/img] THE CALCULATION WILL WORK WITHOUT THIS BEING DONE BUT WILL LOOK HORRIBLE IF YOU DON'T KNOW WHAT YOU ARE DOING!

The final step will be to enable photon mapping from the Brazil: Photon Server rollout in the render dialog
[img]http://www.monstaronline.com/_uploaded/brazil_luma_rollout.jpg[/img].
THE CALCULATION WILL NOT WORK WITHOUT THIS DONE!  I won't go into detail about the parameters of this rollout because the
documentation that splutterfish provides with brazil with the purchase is enough.

For a great tutorial in Brazil r/s on photon mapping [url=http://www.splutterfish.com/sf/contrib/tutorials/pixelforge_photons/_contents/]CLICK HERE[/url].

Alot of people tend to think that the type of renderer is just as important
as the type of software to use when actually the renderer has as much influence in your final render as it does the price of rice in china.   If you are using an advanced 3d application and know what you are doing and everything is to scale, you can take the same concept and apply it to whatever your environment is.  As long as the application you are using supports Radiosity (or at least the ability to fake radiosity with Indirect Illumination).

If you feel up to it, google some document proposals from previous Siggraph conventions.  These are usually in PDF format
and are very educational (if you can get through all the mathematical equations).  Without events like Siggraph, there would be
no Pixar, there would be no ILM, there would be no Blur studios, there would be no computer graphics in the movies and games we play today,
 there wouldn't even be this conversation here.
I hope that this informational post (not tutorial lol) has helped anyone who was having problems with indoor and/or physically correct lighting scenarios.
I'd love to see some renders if anyone decides to try it out.  Don't worry, it will look crappy before it looks nice so keep at it.  Took me a while to pick it up but once you do it will change the way that you render things.

Also, Google Wann Jenson's excellent book, Realistic Image Synthesis Using Photon Mapping if you really want to get into
physically correct lighting.

And make sure you check out my latest .NET application at http://www.ladymonstar.com [/quote]

Disable smileys?
Use signature?
Turn BBCode off?
Receive email notification of new replies?

Topic Review
This is a long topic, click here to review it.