Face Reconstruction from Skull Based on Partial Least Squares Regression

Craniofacial reconstruction can be applied in many fields such as forensic, archaeology etc. This paper proposes to learn the relationship between the face and the skull by Partial Least Squares Regression (PLSR), and reconstruct a person´s face skin from his skull using the relationship. In order to better represent shape variations of skulls and skins, we divide the skull and skin into five correspondent local feature regions respectively, and five mappings from the skull regions to the corresponding skin regions are obtained by PLSR. For an unknown skull, we can get five skin regions by the five mappings learned, and the face can be recovered by stitching the five skin regions. Experimental results validate the proposed method.