Quantization-free parameter space reduction in ellipse detection

Abstract

A new ellipse detection algorithm based on Hough transform and Direct Least Square method is presented. Pair of edges equidistant to ellipse center are used to determine ellipse center location. Normal distribution is used to model the presence probability of ellipse edges from ellipse center. Only edges pertinent to ellipse are passed onto Direct Least Square method for ellipse fitting in order to determine the rest of ellipse parameters. The average time complexity of the algorithm is cO(n2) where c is a fraction. The behavior and merits of the proposed algorithm are shown through several applications.