SuperDSM is a globally optimal segmentation method based on superadditivity and deformable shape models for cell nuclei in fluorescence microscopy images and beyond.