3D Face Recognition at LIVE
The 3D Face Recognition research at LIVE is being conducted in collaboration with former Advanced Digital Imaging Research, LLC, Friendswood, TX.
Introduction to 3D Face Recognition
- Automated human face recognition is a non-trivial computer vision problem of considerable practical significance. It has numerous applications including automated secured access to ATM machines and buildings, automatic surveillance, forensic analysis, fast retrieval of records from databases in police departments, automatic identification of patients in hospitals, checking for fraud or identity theft, and human-computer interaction. Considerable research attention has been directed, over the past few decades, towards developing reliable automatic face recognition systems that use two dimensional (2D) facial images. Commercial systems are also now available for 2D face recognition. Two dimensional face recognition systems are inadequate for robust face recognition. They are known to perform poorly for facial images with uncontrolled illumination or poses.
- Three dimensional (3D) face recognition technology is now emerging, in part, due to the availability of improved 3D imaging devices and processing algorithms. For such techniques, 3D images of the facial surface are acquired using 3D acquisition devices and are used for recognition purposes. Three dimensional facial images have some advantages over 2D facial images. Their pose can be easily corrected by rigid rotations in 3D space. The shape of a 3D facial surface depends on its underlying anatomical structure. Hence, images acquired using 3D laser range finders are invariant to illumination conditions during image acquisition. Three dimensional facial images also provide structural information about the face (e.g., surface curvature and geodesic distances), which cannot be obtained from a single 2D image.
- The field of 3D face recognition deals with the development of algorithms for the (a) identification, and (b) authentication of human beings using their 3D facial models.
- Authentication: This is a one-to-one matching task, wherein a person claims to be a specific entity known to the system. The database of people known to the system is referred to as the 'gallery'. The individual whose identity is verified/authenticated by the system is referred to as a 'probe'. A facial representation of the probe is captured in real-time and compared against the gallery representation of the claimed entity. If the similarity score between the two is greater than a predefined threshold, the individual is verified as the claimed entity; otherwise, he or she is rejected as an imposter. An example of a verification scenario is a system for automated secured access to a building.
- Idenification: This is a one-to-many matching task wherein an unknown individual's identity is established by comparing his/her probe face against a gallery of faces of known individuals. The closest matches in the gallery are found and ranked in descending order of their similarity scores. The probe is assigned the identity of its closest matched face in the gallery.
The 3D Face Recognition web pages are being maintained by Shalini Gupta