In this article we present an algorithm for the detection of perceptually relevant alignments of points in 2D point patterns. Our algorithm is based on the a contrario detection theory. It requires no parameter tuning and has only one critical parameter, which controls the number of false detections.