This article presents a numerical method for facial reconstruction. The problem is the following: if I only have a dry skull, can I reconstruct a virtual face that would enhance the identification of the subject? Our approach combines classical features as the use of a skulls/faces database to learn the relations between the two items and more original aspects: (i) we use an original shape matching method to link the unknown skull to the database templates; (ii) the final face is seen as an elastic 3d mask which is adapted onto the unknown skull. Using our method the skull is considered as a whole surface-represented by a surface triangulation-and not restricted to some anatomical landmarks, allowing a dense description of the skull/face relationship. In particular, our approach is fully-automated. We present some preliminary results to show its efficiency.