An efficient curve detection algorithm

This paper presents a highly-efficient algorithm for the detection of lines, circles, and ellipses. The algorithm uses a novel method for the determination of candidate curves, and no accumulator is needed to save the information of the related parameters. We first define a curve parameter by random sampling and then calculate the amount of points on the curve after selecting several test points in the test area. If the amount of points on this curve is larger than a certain threshold, the curve is then considered as a candidate curve. Afterwards, the evidence-collecting process will be used to further affirm whether this candidate curve is true or not. A large number of experiments have shown that this new method has the advantage of a faster speed of about one to two orders of magnitude compared with the randomized Hough transform algorithm. Besides, it also has a higher detection accuracy and strong robustness.