Most edge detection methods are based on first-order or second-order differential. These are local methods. Using Hausdorff distance to quantify the strength of the edge is a method with a holistic property. Firstly, down sample the image, and split the image into two sets. Secondly, get the feature image by assigning a value for each point using the scalar field map constructed by Hausdorff distance. Since the edge features have local properties, in this paper, we constructed a map which can get local feature images using sub image and combined them into a feature image. Finally we present a method to get the edge image by feature image. We experimentally verify the feasibility of this method.