A threshold strategy for edge tracking

In this paper, we address a threshold strategy based on fuzzy sets theory, which can be used to guide an edge tracking execution. First, we briefly introduce the principle of edge tracking technique. A set of edge models is proposed. The cost functions are defined not only according to the change of local gradient magnitude but also to the change of local gradient directions. A set of thresholds is defined in terms of the histogram of edge enhanced images to determine the start and the end of edge tracking. So that, no matter how much the intensity levels of images vary, the algorithm will always gain the best edge extraction. The fuzzy sets concept is used to combine the human linguistic language into the operation of the determination of edge contrast to the background. A more acceptable interface is established according to the human visual perception. The new algorithm has been tested on several real world images and the experimental results are given.