We propose a method for detecting geometric structures in an image, without
any a priori information. Roughly speaking, we say that an observed geomet
ric event is "meaningful" if the expectation of its occurences would be ver
y small in a random image. We discuss the apories of this definition, solve
several of them by introducing "maximal meaningful events" and analyzing t
heir structure. This methodology is applied to the detection of alignments
in images.