Among the many studies that have appeared in literature, for detecting damage with modal data, methods based on natural frequencies changes have the principal attraction of making use of a reduced set of experimental data that is also easily measurable and less contaminated by experimental noise. However, dynamical shapes have the undoubted advantage of possessing the geometry of the system as their inherent feature and, therefore, damage can be potentially determined by processing the relevant geometric changes of the shapes. In this perspective the data should be extracted from the system under investigation without recurring to numerical models that try to emulate the real system. Such processing techniques that are applied on discrete data (mode shapes or equivalent dynamical shapes) are herein indicated as digital shapes processing techniques in accordance with DSP-techniques. The present article reviews recent and past proposals that aim at processing shapes in order to detect localised damage. The performances of those techniques and the factors that might limit its successful applications are summarized by also processing experimental data.