An Improved Hough Transform for Circle Detection using Circular Inscribed Direct Triangle
Qiang Li, Mingyun Wu · 2020
Considering that the classical Hough transform for circle detection has poor real-time performance because of massive computation and memory consumption caused by accumulative voting in three-dimensional parameter space, this paper presents an improved Hough circle detection algorithm using circular inscribed direct triangle. The algorithm reduces the three-dimensional parameter space needed by classical Hough transform to two-dimensional parameter space by utilizing the geometrical characteristics of circle. By traversing an s pixels only once, the coordinate points which may be centers of candidate circles are voted, and the radius corresponding to the coordinate points whose number of votes exceeds the user-defined threshold is taken out. Finally, the false circles are eliminated, and all real circles in the image are detected. The experimental results show the algorithm has fast speed, high accuracy and good anti-noise performance in contrast with the classical Hough circle detection method.