The aim of the multi-view 3D face descriptor is to recognize a face in any view. The basic idea is based on pose-mosaic and descriptor optimization with new quasi-view concept.