The importance of genetic algorithms in computer art
It is easy to produce simple algorithms
that will give interesting images but, as in this case, the number of
possible images is so huge that
the best way is to let the computer generate them randomly. Unfortunately
trying to get a good image in this way is like waiting for the proverbial
ape to produce a Shakespeare sonnet by clicking randomly on a keyboard.
To accellerate the process of finding these interesting images, the genetic
algorithm is fundamental. This method has been explored from computer artists
like William Latham ,Roman Verosko and Karl Sims (in fact the basic idea behind the images in
the gallery was described from K.Sims at Siggraph 91). The idea behind genetic
algorithms is very simple
In the end the essential ingredients are:
In this way the artist will drive the evolution of new images without having to know
any detail on the maths behind.
An alphabet of a dozen of "letters" .In my gallery these are 16.
- Any word with letters from this alphabet can be interpreted to draw an image(in my
case each "letter" in the word is used to compute a new color of a pixel
in the image by applying a simple function like sin(x) to the old value
of the pixel).
- The user can select interesting images.
- The computer will mutate randomly the words producing these images getting
possibly even more interesting images(this is called random mutation and
consists in changing a random letter in the word representing the image).
- The computer will breed selected words (i.e. cut them in some way and
pasting the pieces to get new words that will produce other images).
This is also called Evolutionary Art.The artist applies the
principles of Darwinian Evolution to breed genetic algorithms from gene banks.
The evolutionary art of Steven Rooke explains in details how this technique is applied from an evolutionary artist.
This idea is used also to create images based on the preferences of
Internet surfers like in this
gallery of random art by Andrey Bauer where you can find also a good
description of his method in How random art works.(His method,I think,is exactly the one
described in the Siggraph paper).
You can even experiment with genetic art yourself in this applet where IFS fractals are used to produce the images.
BACK to Explorations in computer graphics
Maintained by Giuseppe Zito: