A simple robust statistical edge detection method is proposed. The proposed method is developed by combining the good points of Kundu’s method, in which the region segmentation of local area is used, and Fukui’s method, in which a statistical evaluation referred to as “separability” is used for edge extraction. We develop a simple new region segmentation method of local areas and a new statistical evaluation for edge extraction and extract a set of edge features, referred to as a “virtual edge”, consisting of separability, estimated edge location, edge orientation, distance between centers of segmented sub-regions, and difference in mean color of sub-regions. The proposed algorithm is successfully applied to many color images, and the results show that some problems of Kundu’s, and Fukui’s methods are solved.