An effective salient edge detection method based on point flow with phase congruency

This paper aims to detect the image edges using the point flow method based on the fusion of multi-scale phase congruency. The vector field of the original point flow method is built according to the image gradient, which is sensitive to noise and cannot distinguish weak edges, making the model fail to provide complete boundaries in the complex images. In this paper, we propose to build the vector field based on the phase congruency, which is an illumination and contrast-invariant feature for describing the image edges and corners in an image. Moreover, the multi-scale phase congruency is used to construct the vector field for the point flow method. We test our method on the BSDS500 dataset and compare it with several classical and advanced edge detectors. The F1 score and figure of merit (FOM) are used to evaluate the performance quantitatively. These two measurements are widely used analytical parameters to characterize the performance of the edge detectors. Experimental results demonstrate that the point flow method with phase congruency has a significant advantage in the salient edge detection in terms of the evaluation performance and the visual effect.