Abstract: Introducing the symmetry of circular operator, an improved morphological detection algorithm is proposed for binary images to avoid the rotation of structural elements so as to raise the detection efficiency. Then, an improved method for EM image segmentation is proposed for gray and color images to enhance the regional features after segmentation, thus providing more reasonable conditions morphologically for the corner detection of gray and color images. Experimental results verified that the method can get better detection result even if the corner detection is done in a noisy or complex background.