This article (in French) comes from my master thesis at the LIP laboratory (Lyon, France) under the direction of Jacques Mazoyer and Bruno Durand. The report is full of colourfull figures. I included the so-called "colour" version of my report, using a hack in PostScript that drew ink/coffee spots randomly on each impression (the PDF version will not change).
A dual-vision of cellular automata is that rather than being active agents, they are the medium upon which signals do propagate (a signal keeps going in a straight line, one or more being generated at computable angles only when two signals meet, some tricks allowing to do "loops of signals").
The results contained in the article are interesting and form the basis for my later work with Véronique Terrier. I could for example build in dimension 2 a signal that could not be built in dimension 1, that is, a signal of speed 
.
Quoting this publication
@MastersThesis{dubacqdea95,
author = {Jean-Christophe Dubacq},
title = {Signaux rapides en plusieurs dimensions},
school = {\'Ecole normale sup\'erieure de Lyon},
year = {1995},
type = {DEA / Master degree}},
month = sep,
note = {LIP, ENS Lyon}
}