01.jpgBiometric technology allows for the verification of an individual’s identity via
parameters such as their fingerprints, iris, voice, DNA ... or facial features.
However, given that most people’s faces have so much in common with one
another (two eyes, a nose, etc.), it’s sometimes difficult for biometric systems
to tell them apart based on flat two-dimensional images. With that in mind, researchers from Florida AtlanticUniversity (FAU) in Boca Raton have
created a computer algorithm that is capable of creating 3D models of faces
based on 2D images.
The system starts with a generic 3D “synthetic face,” the basic parameters of
which are obtained from anumber of 3D images of actual human faces. When
a 2D image of a specific person is presented, the computer projects the synthetic
face onto that flat image plane. The algorithm, compensating for the pose
and illumination of the 2D face, transfers the subject’s features onto the
synthetic face, to create a 3D approximation.
The FAU researchers believe that the technology could be applied not
only to photographs but also to CCTV security footage, for use in criminal
investigations or searches for missing persons. They have also put forward
the rather creepy suggestion that it could be used to create animated 3D
visages of celebrities from the past – hopefully we won’t be seeing Martin
Luther King trying to sell us light beer anytime soon.
