a) Illustration of membership functions u_ij for AR and QS feature vectors x_j in the case of a large value of η_AR. Because of the larger spread in intensity values of the AR, small values of x_j may have a larger AR membership than QS membership. b) EIT 195 Å image from January 1, 2000 with limb brightness correction. c) Segmentation using original PCM algorithm: The darkest parts are classified as active regions. d) Segmentation using PCM with constraints on η_i-values.