1. A minimal archetypical algorithm. For example circle
equals square over star shape. I may be wrong here but
I
was thinking that algorithms are the same as equations
and inside of an equation, a variable like x is sort of like
saying "some number", so there could
be a variable
equation that was sort of like saying "some equation".
So
this would be an archetypical, fantastical and
minimalistic
equation that just sort of suggests a category of
equations
without defining them rigorously. But actually that
version
of this idea only occurred to me after I wrote the title
down for the following idea:

2. an algorithm in a 3D
modeling program that allowed you to exagerate
different
parts of a 3D body shape, smoothly, so that you could
take
an exact scan of a person's body and exagerate the
features smoothly in the same way political cartoonists
exagerate features, but so that the 3D version of this
photo realitic scan could then move around the screen
like
in one of those 3D movies. Maybe they do this already.
I
haven't seen too many movies lately and it kind of
stands
to reason that they would have done this. But I have not
seen anything like that. Actually I guess MASK was sort
of
like that and that was like 10 years ago. But I still
cannot
think of a movie that has a sort of fun house mirror style
distortion of one character walking around amidst a lot
of
regular looking characters.

idea 2 is actually baked. It's very hard to link a bunch of vertexes (3d models) or high resolution scan of someone (point clouds) into "ear" "eye" "nose" as a computer doesn't know how to define these objectives from a 3d shape. The human brain is very adept at picking out the shape of humans and human faces, computers are not.

However there's a trick, by using a library of 100's of faces 3d scanned you can make a stastical model of what a face is. Then you can back solve using a "best fit" type of algorithm basically you look for the smallest difference between your new face and the statistical average of all faces combined.

Finally by using sorted or weighted data sets, you can generate exaggerations or expressions based on the statistical model.

So far all the solutions I've seen are rather brute force in that they don't simulate or predict muscle movement or skeletal frame. They just use the various statistics from a sample set. Therefore some of the facial expressions look derived.

Anyways see link. It's ground breaking stuff, and in the future it will make picture faking and 3d model generation that much easier.