A Fast Randomized Detection Algorithm for Multi-Ellipse Detection Problem

This paper proposes a fast randomized detection algorithm for multi-ellipse detection. The algorithm utilizes one randomly sampled edge point and two locally searched edge points from the image, along with their neighborhood information to generate candidate ellipses. Candidate ellipses are then transformed into corresponding circles, with true ellipses being verified through circle confirmation. The approach minimizes random sampling points while effectively filtering out non-elliptical points, reducing ineffective sampling and computational overhead. Numerical experiments demonstrate the algorithm's strong robustness and significantly faster detection speed compared to similar algorithms.