Abstract: In order to construct normalized 3D face database which knowledge-based morphable model requires, a texture map based on data structure and a point-based segmentation method for 3D face are presented. This structure represents explicitly the implied information of original data, and constructs bidirectional-mapping between the separated sampling points and texture of the real 3D faces scanned by laser scanner. So it can ensure that the two parts of data are processed synchronously and real-time. Using rapid mapping feature of this data structure, the presented segmentation method can process plenty of original data rapidly.