A simple robust statistical edge detection method is proposed. This method solves the following problems of conventional methods: (1) the estimation of the edge existence probability is needed for all pixels, (2) the estimation calculated at a small area separated by a straight line must be calculated in every line direction, (3) a search is needed to find the local maximum evaluation value, and (4) the evaluation value is blurred when the boundary consists of a curved line or texture-like line. The proposed algorithm is applied to many natural images and the results show that the main edges can be extracted by a set of fixed parameters.