Symmetric polarity in generalized symmetry transformation

Generalized symmetry transformation (GST) can detect symmetric objects under the edge constraints of direction and distance. GST is, however, sensitive to random noise since it does not regard symmetric polarities. In this paper, a bipolar edge direction constraint is proposed to obtain robustness to a noisy or complex background by introducing symmetric polarities. The capability of the novel approach is evaluated by experiments on synthetic and real images including noisy or complex backgrounds.