Gestaltic Grouping of Line Segments

Abstract

Using simple grouping rules in Gestalt theory, one may detect higher level features (geometric structures) in an image from elementary features. By recursive grouping of already detected geometric structures a bottom-up pyramid could be built that extracts increasingly complex geometric features from the input image. Taking advantage of the (recent) advances in reliable line segment detectors, in this paper, we propose three feature detectors along with their corresponding detailed algorithms that constitute one step up in this pyramid. For any digital image, our unsupervised algorithm computes three classic Gestalts from the set of predetected line segments: good continuations, non-local alignments, and bars. The methodology is based on a common stochastic a contrario model yielding three simple detection formulas, characterized by their number of false alarms. This detection algorithm is illustrated on several digital images.