Edge linking using geodesic distance and neighborhood information

This paper deals with the edge linking problem. We propose two improvements to existing algorithms. First we propose the use of an application-specific local neighborhood within which to compute edge direction in order to improve the accuracy of the edge direction. Second, we propose the use of geodesic distance for measuring the proximity between two candidate edge points to be linked, so that the intensity image information can be taken into account. With these two improvements, our edge linking algorithm can make better edge linking decisions than existing algorithms, as demonstrated in the experimental results.