A crucial aspect of the discretization of discrete surfaces is related to the surface projection. The simple and efficient algorithms available for the projection to parametric surfaces (see Table 2) cannot be adopted simply because of the missing parameterization of the limit surface. Moreover, the situation is further complicated by the fact that the normal vector of the limit surface can be exactly evaluated only at the nodes of the original or refined control grid. Therefore in order to make the projection sufficiently accurate (in terms of both the distance from the limit surface and the match with the exact normal), it is necessary to subdivide the control grid up to a high level. This results in a huge amount of data to be stored, which is not acceptable.