An algorithm for edge detection by morphological gradient

This paper introduces a nonlinear edge detector based on the morphological gradient, which is extended from the classical morphological gradient in two ways. Edge direction is estimated based on moment calculation; non-maximum suppression is done along the edge normal direction. Because local features (the morphological gradient) and global properties (the moment of larger scale) are considered, all available edge information is explored. A theoretical analysis of how morphological operators affect the edge orientation is discussed. Experimental images and comparisons with other algorithms show that the performance of this algorithm is a reasonable edge detector, and has the benefit of simple implementation.