A computational approach to curve detection in 3D dot patterns is developed in the article. A central theme of the approach is to identify curvilinear clusters based on the perceptual organization of the dots. The main idea is to use some characteristics of the Voronoï diagram of the dots. These characteristics are so called eccentricity, distance measure, and squeezedness. The authors gave an algorithm how to distinct the points lying on “curve-clusters" from other ones, based on these characteristics. The experimental results seem to be impressive. As a weak point of the article I want to mention that the precise definitions of the eccentricity, distance measure, and squeezedness are not presented in the article (they are given only on an intuitive level).
Reviewer: A. Ivanov (Moskva)