3D digital Forensic Facial Reconstruction using Various Human Facial Templates

Dalia Abdou (William Harvey Research Institute, Queen Mary University of London)
Dr. Maria Vanezis (Academy of Forensic Medical Sciences)
Professor Atholl Johnston (William Harvey Research Institute, Queen Mary University of London)
Professor Peter Vanezis (William Harvey Research Institute, Queen Mary University of London)

In three-dimensional Forensic Facial Reconstruction, faces are either reconstructed from “outside inwards” by ‘wearing’ a face template onto the skull (Quatrehomme et al., 1997; Vanezis, 2008), or from “inside outwards” by building the individual muscles from the bone surface outwards (Wilkinson, 2006; Lee et al., 2012). In this pilot study, the first approach was adopted aiming at comparing single human faces to average faces, produced by merging a number of single faces, as facial templates for 3D Forensic facial reconstruction. Two skulls (~30y and 20y) were reconstructed using each of 13 templates (10 single & 3 average). Face pool and resemblance tests were conducted for each facial reconstruction. Facial reconstructions were then ranked according to the results of both test types and agreement among ranks was sought. In face pool tests, all 3 average and only 1 single face have repeatedly received above chance correct identification percentage in both cases. In face resemblance tests, the average 30y and 20y faces received the first ranks for the 2 cases respectively. Average faces are more useful in identification due to their high robustness as templates for different skulls as well as the lower chance of misidentification than single human faces.