Circle Detection Based on Arc Search Using a Table of Virtual Circle
Makoto Odagiri, Kazunori Onoguchi · 2016
This paper proposes the novel method which detect circles in an image by an approach of the template matching, not an approach of the Circle Hough Transform(CHT) voting edge points to the parameter space. The approach such as the Hough transform needs huge computing cost and huge memory. Furthermore, a lot of false circles are detected in a complicated background. In order to overcome these problems, the proposed method detects a circle by examining edge gradient information, such as edge gradient magnitude and edge gradient direction, on a virtual circle created from a pair of edge existing on the horizontal or vertical scanning line. Since a virtual circle is created from a table data which was prepared beforehand, the computing cost can be reduced. In addition, the proposed method is suitable for parallel processing on the multi- processor. Experimental results compared with some conventional methods show the effectiveness of the proposed method.