Fast ellipse detection based on three point algorithm with edge angle information

Abstract

In this paper, we introduce a fast ellipse detection method that uses the geometric properties of three points on an ellipse. Many conventional ellipse detection methods carry out detection using five points, but a random selection of such points among candidate edges requires much redundant processing. To search for an ellipse with the minimum number of points, this study used the normal and differential equations of an ellipse, which requires three points based on their locations and edge angles. First, to reduce the number of candidate edges, the edges were divided into 8 groups depending on the edge angle, and then a new geometric constraint called the quadrant condition was introduced to reduce noisy candidate edges. Clustering was employed to find prominent candidates in the space of a few ellipse parameters. Experiments using many real images showed that the proposed method satisfies both reliability and computing speed for ellipse detection.