Sequential edge linking method for segmentation of remotely sensed imagery based on heuristic search

An improved edge linking method for segmentation of remotely sensed imagery is proposed in this paper. Firstly, the gradient strength and the gradient direction are calculated. Secondly, the initial edge points are generated by the proposed locally maximal gradient strength strategy. Finally, edges are generated by the heuristic search method (A*) which links all the initial edge points based on the gradient strength, gradient direction and edge direction. Most of the existing edge-based segmentation methods would produce discontinuous edges, while the proposed method can produce the segmentation result with precise, continuous, and one-pixel-wide edges. In the experiment, the method is tested successfully on the QuickBird image and performs well.