Skip to Main Content
The morphological gradient is an attractive option for edge detection in applications where morphological filtering is used, but the basic well known implementation has three problems. It provides only a magnitude response without edge orientation and it has a magnitude response which is dependent on the orientation of the object edge. A third problem arises because the morphological gradient is more sensitive to added noise than well known linear gradient estimators such as the Sobel operator, the Laplacian, or the Robert's cross. In this paper, we propose a modified morphological gradient which can overcome the drawbacks discussed above without substantial increase in computation. The performance of this new method is analyzed based on both theoretical and simulation results. Theoretical analysis and experiment show that our method also can be extended to scale space edge detection. Comparisons with other edge detectors use both synthetic images and real images.