Department of Computing Sciences
Villanova University, Villanova, PA 19085
CSC 3990 – Computing Research Topics
Biometrics is the automated identification of a person based on physical traits. One biometric which has received considerable attention in recent years is face recognition. Face recognition is considered to be one of the most challenging biometrics because it depends on variations in image quality, orientation, and the subject’s appearance. This paper discusses current implementations using 2D or 3D based recognition. 2D recognition achieves generally impressive results. However, accuracy decreases drastically when the images being compared have significant variations. Currently, there is much research being done in the area of 3D recognition which hopes to improve upon the inherent limitations of 2D recognition.
Face recognition is an attractive biometric for use in security applications. Face recognition is non-intrusive, it can be performed without the subject’s knowing. This has become particularly important in modern times because demand for enhanced security is in public interest.
2. Facial Recognition Approaches
2.1 Eigenface-based Recognition
2D face recognition using eigenfaces is one of the oldest types of face recognition. Turk and Pentland published the groundbreaking “Face Recognition Using Eigenfaces” in 1991 . The method works by analyzing face images and computing eigenfaces, which are faces composed of eigenvectors. Results obtained by comparing eigenfaces are used to identify the presence of a face and its identity.
There is a five step process involved in the system developed by Turk and Pentland. First, the system needs to be initialized by feeding it a training set of face images. These are used to define the face space which is a set of images that are face-like. Next, when a face is encountered, the system calculates an eigenface for it. By comparing it with known faces and using some statistical analysis, it can be determined whether the image presented is a face at all. Then, if an image is determined to be a face, the system will determine whether it knows the identity of the face or not. The optional final step concerns frequently encountered, unknown faces, .which the system can learn to recognize.
The eigenface technique is simple, efficient, and yields generally good results in controlled circumstances . The system was even tested to track faces on film. However, there are some limitations of eigenfaces. There is limited robustness to changes in lighting, angle, and distance . Also, it has been shown that 2D recognition systems do not capture the actual size of the face, which is a fundamental problem . These limits affect the technique’s application with security cameras because frontal shots and consistent lighting cannot be relied upon.
2.2 3D Face Recognition
3D face recognition is expected to be robust to the types of issues that plague 2D systems . 3D systems generate 3D models of faces and compare them. These systems are more accurate because they capture the actual shape of faces. Skin texture analysis can be used in conjunction with face recognition to improve accuracy by 20 to 25 percent . The acquisition of 3D data is one of the main problems for 3D systems.
2.3 How Humans Perform Face Recognition
It is important for researchers to know the results of studies on human face recognition . This information may help them develop ground breaking new methods. After all, rivaling and surpassing the ability of humans is the key goal of computer face recognition research. The key results of a 2006 paper “Face Recognition by Humans: Nineteen Results All Computer Vision Researchers Should Know About”  are as follows:
Humans can recognize familiar faces in very low-resolution images.
The ability to tolerate degradations increases with familiarity.
High-frequency information by itself is insufficient for good face recognition performance.
Facial features are processed holistically.
Of the different facial features, eyebrows are among the most important for recognition.
The important configural relationships appear to be independent across the width and height dimensions.
Face-shape appears to be encoded in a slightly caricatured manner.
Prolonged face viewing can lead to high level aftereffects, which suggest prototype-based encoding. See Figure 1 for an example
Figure 1. Staring at the faces in the green circles will cause one to misidentify the central face with the faces circled in red .
Pigmentation cues are at least as important as shape cues.
Color cues play a significant role, especially when shape cues are degraded.
Contrast polarity inversion dramatically impairs recognition performance, possibly due to compromised ability to use pigmentation cues. See Figure 2.