Exploiting 3D faces in biometric forensic recognition

The analysis of 3D face data is very promising in improving the identification performances of individuals. In this paper two algorithms will be presented to match 3D face shapes (3D to 3D) and also to compare a synthetically generated 2D view with a given 2D face image (3D to 2D). The algorithm to match 3D shapes is based on the extraction and comparison of invariant features from the 3D face template. In order to perform the 3D to 2D matching an algorithm is used to generate a 2D view of the face which is as close as possible to the given face image. Salient features are extracted from the two face images and are used to determine the similarity of the faces. To simulate a real forensic environment, a small database of 3D faces has been acquired. The dataset will be used to test the identification performances in both cases.