CDR: Gort in granny-shades (was Re: Al Gore goes cypherpunk?)

[David Honig]

At 08:08 PM 10/24/00 -0400, Tim May wrote:
>
>Nonsense, on at least a couple of accounts. I was active in the image 
>processing field in 1980-84, and attended various SIGGRAPHs and 
>suchlike. Fact is, "ray tracing" and various illumination models, and 
>Gouraud and Phong shading and all the rest...were NOT motivated by a 
>desire to model "*physics*." Physicist didn't give a dang about 
>modelling light sources in 3D environments, and about morphing and 
>wrapping and all that.

Not *that* kind of physicist.  Something more like psychophysics (ie
measurement of human perception) empowered by these newfangled computer
thangs.

There was a desire to see what was sufficient to generate photorealism.
After all, the shading models you mention are crude approximations; but
they usually work[1], which talks to the limits of perception.  And
to a nontrivial physics of a certain scale (modelling translucent
objects, objects with embedded reflectors and adsorbers, etc.)

There was also a desire to understand the rendering process so that
you could understand the *inverse*, ie, inferring the world from
what you see.  Machine vision.

Rendered jello, fires, fractal cloud effects, 
realistic water, rocks, etc.  Hollywood was for a while making entire
movies around a single new model (particle systems in that terraforming
star trek movie).  The question that motivates was, how do you do that?
Where 'that' might be rendering hair, or modelling how skin creases at elbows.

I modelled wood grain with a VAX in 1985 using a 24 bit monitor 
that cost more than some cars.  Nowadays you would just scan and
texture map real wood, but the (psychophysical) question was,
what did you need to model to get grain indistinguishable from real[2]? 
Texture mapping is cheating.

Similar questions exist for modelling motion.  How many harmonics
do you need to render for realistic motion?  How do you make a 
desklamp move with emotion?  (Luxo Jr..)  (Recording actors and
mapping movement is cheating, Mr. Spielberg.)  

>The motivation was to produce special effects for education films (a 
>la James Blinn at JPL), effects for movies (a la Alvy Ray Smith, 
>eventually of Pixar), and advertisements for Hollywood and Madison 
>Avenue.

These folks had a more academic interest than you make out.. Blinn wrote a
very mathematical column.. the sci films and movies paid the bills, bought
the equiptment, etc.,  and sometimes motivated problems, but the problems
were fascinating in and of themselves.  Sure, their output was often
visual candy, but it was new and interesting regardless of jazz appeal.


[1] The moon does not show the cos(incidence-angle) reflectance
of typical matte surfaces, and thus looks flat.  Its surface is powder.

[2] slice vertical concentric shells with radially varying albedo (annual
rings) paired with small vertical radial planes (rays).  The 3D geometry
yields 2D constraints on the texture that distinguish it from other
striated textures.  Turning the knobs generates different types of 
woodgrains.