In this paper, a new method for edge detection in multispectral images is presented. The method is based on the assumption that edges correspond to the boundary that separates two adjacent homogeneous regions, and characterization of edges are conducted by evaluating the difference between the two adjacent neighborhood regions to derive edge intensity and orientation. To demonstrate the efficiency of the proposed approach, we carried out computer experiments on multiple-band images and comparison with other methods. Experiments on a variety of images have shown that it is consistent and reliable tool to detecting edges of multispectral images. The proposed method can detect the edges successfully and can resolve to thick edges.