A computerized craniofacial reconstruction method for an unidentified skull based on statistical shape models

Craniofacial reconstruction (CFR) has been widely used to produce the facial appearance of an unidentified skull in the realm of forensic science. Many studies have indicated that the computerized CFR approach is fast, flexible, consistent and objective in comparison to the traditional manual CFR approach. This paper presents a computerized CFR system called CFRTools, which features a CFR method based on a statistical shape model (SSM) of living human head models. Given an unidentified skull, a geometrically-similar template skull is chosen as a template, and a non-registration method is used to improve the accuracy of the construction of dense corresponding vertices through the alignment of the template and the unidentified skull. Generalized Procrustes analysis (GPA) and principal component analysis (PCA) are carried out to construct the skull and face SSMs. The sex of the unidentified skull is then predicted based on skull SSM and centroid size, rather than geometric measurements based on anatomical landmarks. Furthermore, a craniofacial morphological relationship which is learnt from the principal component (PC) scores of the skull and face dataset is used to produce a possible reconstructed face. Finally, multiple possible reconstructed faces for the same skull can further be recreated based on adjusting the PC coefficients. The experimental results show that the average rate of sex classification is 97.14% and the reconstructed face of the unidentified skull can be produced. In addition, experts’ understanding and experience can be harnessed in production of face variations for the same skull, which can further be used as a reference for portraiture creation.