Evolutionary image registration in craniofacial superimposition: Modeling and incorporating expert knowledge

Abstract

Craniofacial superimposition involves the process of overlaying a skull with a number of ante-mortem images of an individual and the analysis of their morphological correspondence. This research focused on the skull-face overlay stage with the aim of modeling the expert knowledge that is related to the existing anthropometric differences among landmarks and incorporating it into this stage. Consequently, we have moved from a single-objective optimization problem to a multiobjective optimization one aimed to reduce the distances between pairs of landmarks from each group independently. To tackle it, two classic approaches from the area of multicriteria decision making were used: weighted sum and lexicographical order. The results, which were obtained over a Ground Truth dataset, are promising in those cases where the forensic expert has located a large number of landmarks, and worse results than the state of the art method in cases with few landmarks.