> http://www.ibm.com/developerworks/java/library/j-j2d/
You seem to have chosen as your CA filters the kind
that reduce eventually to generating slow moving
coherent waves across the base image.
Also interesting might be to use "Game of Life"-like
binary cell-valued CAs (the one called "37/23" is
especially nice IMO), on a toroidal surface, which
GOL-like instance usually stays busy for a very,
very long time without reducing to wavelike
behaviors, starting from a random 50% fill start of
at least some part of the play-field.
Instead, the behavior seems to be to decrease
quickly to a sparse play-field, where much of that
play-field is areas of stability or of simple
repeated cycling, which are occasionally overrun by
areas of fast moving quasi-chaos.
This would give you entities moving across your
background image in bewildering ways.
The tool Golly is good for exploring how such CAs
behave:
Message-Id: <andrew-2810...@192.168.1.2>
alternately
http://preview.tinyurl.com/5epw7k
Using well known variations of that "Game of
Life"-like process where the cell values are still
logically binary, but where for display purposes,
currently unset cells decay to show an age gradation
[for a while; 6 decay steps seems to be a common
choice] of how long since they were last set, gives
you a somewhat broader palette (eight colors vice
two) with which to modify your background image.
What you've done that is _really_ clever, of course,
is to use your CA state as a filter against an
unvarying background image, rather than varying that
background image as itself a CA. That way lies
wonderful stuff, and you've only just begun to seek
out those wonders.
For example, a CA where the rule-set of every cell
was chosen in some way from a large set of possible
rule-sets by some features of the underlying
background pixel would be likely to be at the very
least a form of CA having universal computer
potential, and also a strong form of image analysis
tool.
xanthian, rambling.