Face Recognition Techniques. The Problem of Face Recognition

Introduction to Face Recognition. Biometrics is becoming a buzzword due to increasing demand for user-friendly systems that provide both secure and efficient services. Currently, one needs to remember numbers and/or carry IDs all the time, for example, a badge for entering an office building, a password for computer access, a password for ATM access, and a photo-ID and an airline ticket for air travel.

Although very reliable methods of biometric personal identification exist, e.g., fingerprint analysis and iris scans, these methods rely on the cooperation of the participants, whereas a personal identification system based on analysis of frontal or profile images of the face is often effective without the participant’s cooperation or knowledge.

It is due to this important aspect, and the fact that humans carry out face recognition routinely, that researchers started an investigation into the problem of machine perception of human faces. In Fig. 1, we illustrate the face recognition task of which the important first step of detecting facial regions from a given image is shown in Fig. 2.

Figure 1. An illustration of the face recognition task (1): given an input facial image (left column: many variants of the facial image are used to illustrate image appearance change due to natural variations in lighting and pose, and electronic modifications that simulate more complex variations), matching it against a database of facial images (center column), and finally outputting the matched database image and/or the ID of the input image (right column)

Figure 2. Detection/Segmentation/Recognition of facial regions from an image (2)

After 35 years of investigation by researchers from various disciplines (e.g., engineering, neuroscience, and psychology), face recognition has become one of the most successful applications of image analysis and understanding. One obvious application for face recognition technology (FRT) is law-enforcement. For example, police can set up cameras in public areas to identify suspects by matching their imagaes against a watch-list facial database.

Often, low-quality video and small-size facial images pose significant challenges for these applications. Other interesting commercial applications include intelligent robots that can recognize human subjects and digital cameras that offer automatic focus/exposure based on face detection. Finally, image searching techniques, including those based on facial image analysis, have been the latest trend in the booming Internet search industry. Such a wide range of applications pose a wide range of technical challenges and require an equally wide range of techniques from image processing, analysis, and understanding.

The Problem of Face Recognition. Face perception is a routine task of human perception system, although building a similar robust computer system is still a challenging task. Human recognition processes use a broad spectrum of stimuli, obtained from many, if not all, of the senses (visual, auditory, olfactory, tactile, etc.).

In many situations, contextual knowledge is also applied (e.g., the context plays an important role in recognizing faces in relation to where they are supposed to be located). However, the human brain has its limitations in the total number of persons that it can accurately ‘‘remember.’’ A key advantage of a computer system is its capacity to handle large numbers of facial images.

Figure 3. Configuration of a generic face recognition/processing system. We use a dotted line to indicate cases when both face detection and feature extraction work together to achieve accurate face localization and reliable feature extraction

A general statement of the problem of the machine recognition of faces can be formulated as follows: Given still or video images of a scene, identify or verify one or more persons in the scene using a stored database of faces. Available collateral information, such as race, age, gender, facial expression, or speech, may be used to narrow the search (enhancing recognition).

The solution to the problem involves face detection (recognition/segmentation of face regions from cluttered scenes), feature extraction from the face regions (eyes, nose, mouth, etc.), recognition, or identification (Fig. 3).