Ellipse Detection by Combining Division and Model Selection Based Integration of Edge Points

We present a new method for detecting ellipses in images. We first detect edges, from which we select an elliptic arc by RANSAC: we randomly choose five points repeatedly until the ellipse they define closely passes through a maximum number of edge points. Those points are regarded as constituting an elliptic arc. We repeat this procedure and detect multiple elliptic arcs. Then, we integrate those arcs that define similar ellipses to one group and fit a single ellipse to them. The similarity between ellipses is measured by a model selection criterion. Thus, our method can accurately detect partially occluded ellipses. Using real images, we demonstrate the effectiveness of our method.