Shape understanding and modelling are complex procedures, which are the result of different specific processes. Generally, many steps in these processes may require a high computational-effort, thus making the use of parallel processing an interesting possibility. A typical example is feature recognition, a key tool towards shape understanding, which is applied to detect either local or region-wide characteristics. In this paper a survey is provided about the possibility of using parallel processing at different stages of the shape-understanding process. The parallelisation strategies will be described in details and results will be provided both for workstation networks and massively parallel machine (Cray T3D).