Facial Reconstruction as a Regression Problem

Abstract

This chapter presents a computer-assisted method for facial reconstruction. This method provides an estimation of the facial outlook associated with unidentified skeletal remains. Current computer-assisted methods using a statistical framework rely on a common set of points extracted form the bone and soft-tissue surfaces. Facial reconstruction then attempts to predict the position of the soft-tissue surface points knowing the positions of the bone surface points. This chapter proposes to use linear latent variable regression methods for the prediction (such as Principal Component Regression or Latent Root Root Regression) and to compare the results obtained to those given by the use of statistical shape models. In conjunction, the influence of the number of skull landmarks used was evaluated. Anatomical skull landmarks are completed iteratively by points located upon geodesics linking the anatomical landmarks. They enable artificial augmentation of the number of skull points. Facial landmarks are obtained using a mesh-matching algorithm between a common reference mesh and the individual soft-tissue surface meshes. The proposed method is validated in terms of accuracy, based on a leave-one-out cross-validation test applied on a homogeneous database. Accuracy measures are obtained by computing the distance between the reconstruction and the ground truth. Finally, these results are discussed in regard to current computer-assisted facial reconstruction techniques, including deformation based techniques.