Detecting partially occluded ellipses using the Hough transform

Abstract

This paper considers the problem of detecting elliptical curves using Hough transform (HT) methods. Storage and efficiency problems are overcome by decomposing the problem into two stages. The first stage uses a novel constraint as the basis for a Hough transform to detect the ellipse centre while the second stage finds the remaining parameters using a simple but efficient focusing implementation of the HT. The method is applicable in many situations where previous HT schemes would fail. Results are demonstrated for complicated image data containing several overlapping and occluding ellipses.