Edges are important primitives for any high level visual processes. There must be hundreds, if not thousands, proposed edge detectors. Then why do we want another edge detector developed? The reason is because current edge detectors are sensitive to the presence of noise. If a detector is designed to accommodate noise, then it may become insensitive to edges. Therefore, there is a balance at which the detection of edges and the rejection of noise can be optimized. Previous approaches employ straightforward optimization techniques, which do not work well in real images. We approach this problem based on the evaluation results, which are used to refine the edge detection process. The detected edge is evaluated based on a number of criteria, the results of which are then used to refine the detection process. In fact, our approach is unique because of the incorporation of evaluation as part of the detection mechanism.