From LightWiki

This tutorial was written before the advances in LightWave 9 - especially 9.2 - and should only really be referred to by users of LightWave versions older than that - the management

Well it took me 27min. to "DO" everything in the Street Wise (StreetWise_-_An_intermediate_tut_for_the_road) tutorial the first time thru and it's taken me now nearly two and a half hours to write, check, and grab graphics for this much so far. Anyone wanna guess which is the harder job? Man, next time I see a useful long tutorial online I'm sending flowers AND chocolates!

Snail Race

And they're off! Just look at that render go! Oh wait, I haven't told ya how to do it yet. Likely most people who have the brass to sit there and fiddle around with settings in a logical manner have already discovered this but for those who haven't tried this yet here's a walk thru and brief explanation to ultra fast preview quality radiosity renders. Call it a poor man's FPrime and thus why it has found it's way into this tutorial which earlier on mentioned FPrime considerably.

Also see: http://www.spinquad.com/forums/showthread.php?s=&threadid=245

Basically this technique is achieved by keeping the camera's view plane smaller than the interpolated ray evaluation grid spacing in "Interpolated" Radiosity.

Most people hate the results of interpolated radiosity for the ugly and obvious blotches that the interpolated ray emissions introduce into the GI model. But if the camera view is strictly limited to the span of the interpolated ray matrix grid points which is user definable, and never allowed to view a surface containing more than one interpolated ray emission point then the resulting image appears at least satisfactory for high quality previews and in some few cases good enough for final output.

The main reason that this is interesting are the resulting render times in LightWave3D. With a total maximum of four grid points to express the ray emission for the illumination from any given surface LW3D hardly even notices the load and render times are on par speed wise with scenes that have all ray tracing options turned off. At this point the GI model adds virtually no extra time to the rendering process yet still the "basic" look and feel of radiosity is achieved. Again, achieved to a degree acceptable for HQ preview renders at least.

With careful planning and consideration this effect can be used satisfactorily in animation. The problem with this effect in animation is that the sampled grid is overwritten for each evaluation pass. And to the best of my ability at guessing I conclude that this occurs per frame (or per camera take) in any given animation. The results appear as unsteady flashes or varying intensities of light over time - similar to a strobbing effect. Since it is still hundreds of times faster than Monte Carlo it is possible to oversample many camera takes per frame and average the results to produce a frame with very much less variation from frame to frame and therefore less flashiness.

This method of oversampling for animated sequence frames can be employed in two ways. One is within LW's own rendering engine and is initiated whenever motion blur or like effects are enabled. The other is in an external application such as Digital Fusion (now just plain Fusion) which is capable of evaluating per-pixel intensities across a specified number of frames and averaging the results to compose a supersampled image sequence from a normally rendered animation.

Further, employing both methods in combination almost completely eliminates any strobing visible to the human eye. Certainly far less than that introduced by NTSC interlacing in a common television set!

So in this last case for video, and where scenes can be designed and arranged within the above specified limitations, interpolated radiosity can be considered a viable and advantageous solution!

So let's SEE what all this noise is about, aye?

I won't go into any of the processes required to produce acceptable animation sequences with this technique. I'll save that for the next time I'm inspired to wear out my keyboard.

For now I'll just set up one or two still solutions and compare them with FPrime for your amusement. I fully believe they will not compare favorably but where the user does not have access to FPrime or where the hardware spec is too low this is one alternative solution.

Let's scope out that street we just made.

Here's the LW GI settings panel so you can see how it was set. This panel is from an earlier version of LW 9.2, where the rays were still expressed in a function, and there was no MPS system yet. Functionally things are the same otherwise.

Here's the shot rendered in LightWave's native interpolated radiosity.

Here's the FPrime render given the same 40 seconds.

Here's one more for the road in native LW.

And the equivalent FPrime preview render given 1 min.

Here's a little beefier test for both FPrime and LW using this technique. First here's the GI for LW on this scene.

First the native LW interpolated render followed by the FPrime render equivalent given the same amount of time.