A Prescreener for 3D Face Recognition Using Radial Symmetry and the Hausdorff Fraction

Abstract

Face recognition systems require the ability to efficiently scan an existing database of faces to locate a match for a newly acquired face. The large number of faces in real world databases makes computationally intensive algorithms impractical for scanning entire databases. We propose the use of more efficient algorithms to "prescreen" face databases, determining a limited set of likely matches that can be processed further to identify a match. We use both radial symmetry and shape to extract five features of interest on 3D range images of faces. These facial features determine a very small subset of discriminating points which serve as input to a prescreening algorithm based on a Hausdorff fraction. We show how to compute the Haudorff fraction in linear O(n) time using a range image representation. Our feature extraction and prescreening algorithms are verified using the FRGC v1.0 3D face scan data. Results show 97% of the extracted facial features are within 10 mm or less of manually marked ground truth, and the prescreener has a rank 6 recognition rate of 100%.

Robust realtime feature detection in raw 3D face images

“…They report 82.7% accuracy for 3D faces with neutral expression and 75% for non-neutral expression. Koudelka et al [12] argue that the nose and eyes are radially symmetric and can be detected using the radial symmetry transform [13]. They detect five feature points namely the nose tip, sellion, inner eye corners, and center of the mouth.…”

“…They detect five feature points namely the nose tip, sellion, inner eye corners, and center of the mouth. Using the FRGC v1.0 database, Koudelka et al [12] report that 97% of the extracted features are within 10 mm of the manually marked ground truth. They also preprocess the data for removing spikes and filling holes before feature detection and report 2.5 seconds run time using Matlab on a 3.4 GHz machine.…”

Point-pair descriptors for 3D facial landmark localisation

“…Lu and Jain [7] propose a feature extractor based on the directional maximum to estimate the nose tip location and pose angle simultaneously. Koudelka et al [21] used radial symmetry and shape to extract features of interest on 3D range images of faces. Different pose-dependent and multimodal approaches to localise facial landmarks using the FRGC database have been also reported [3], [10]- [12].…”

Secure portal.

“…Prior to deformation, each input face F is segmented, filtered to remove noise and fill holes, and registered to a scaled version of the reference face using ICP. The reference face was scaled vertically as in [14] using sellion and mouth features that were autonomously detected using radial symmetry [19]. Examples of segmented faces and their resulting deformed faces are shown in Figure 3 for various facial expressions of the same subject.…”