A robust, real-time ellipse detector

In this paper, we present an improved ellipse detector that may be used in real-time face detection. In this algorithm we first extract edges from the image using a robust edge detector, which is then followed by a rule-based method. Finally, we decompose the parameter space of the Hough transform to achieve computational efficiency. Our extensive experimental results show indeed that the proposed detector is capable of detecting ellipse features with an excellent accuracy under various image conditions at a high speed (10 frames per second on a Pentium-III 500MHz PC).