Curve detection using Delaunay triangulation based on perceptual grouping factors
Takamitsu Takagi, Fumihiko Saitoh · Electronics and Communications in Japan · 2012
Abstract In this paper we propose a method of detecting perceptual continuous curves in a binary image including discontinuous curves and noise by means of perceptual grouping factors. First, by making a Delaunay diagram from a binary image, the adjacency relations are obtained. Second, the method selects a Delaunay edge as the starting line and connects a certain number of points. When the points have the maximum evaluation value based on perceptual grouping factors, they construct an appropriate partial curve. The evaluation value of an appropriate partial curve is added to the Delaunay edges selected by it. Finally, Delaunay edges that have high evaluation values are extracted. Experimental results show that the proposed method yields results similar to visually perceived curves and better than conventional methods. © 2012 Wiley Periodicals, Inc. Electron Comm Jpn, 95(7): 19–28, 2012; Published online in Wiley Online Library ( wileyonlinelibrary.com ). DOI 10.1002/ecj.11388