Detection of Symmetrical Objects by Combinational Hough Transform Using Edge Direction and Curvature of Pixel Pair
Hitoshi Terazawa, Fumihiko Saitoh · IEEJ Transactions on Industry Applications · 2003
This paper proposes a method to detect symmetrical objects that have the line symmetrical form and the point symmetrical form in an objective image. When a pair of edge pixels is judged to have the symmetrical relation each other by the geometrical parameters of the directions and the curvatures of the edge pixels, the combinational Hough transform based on the geometrical parameters is applied to vote into the parameter spaces to detect the symmetrical objects. After the voting process about all pairs of edge pixels is completed, the addresses that have large voted values are detected as the locations of symmetrical objects. The experimental results show that the line symmetrical objects and the point symmetrical objects were detected successfully even if an image included both objects.