Biometrical Processing of Faces in Security and Forensics

Biometrical Processing of Faces in Security and Forensics

Pawel T. Puslecki (National University of Ireland, Ireland)
DOI: 10.4018/978-1-60566-836-9.ch004
OnDemand PDF Download:
$30.00
List Price: $37.50

Abstract

The aim of this chapter is the overall and comprehensive description of the machine face processing issue and presentation of its usefulness in security and forensic applications. The chapter overviews the methods of face processing as the field deriving from various disciplines. After a brief introduction to the field, the conclusions concerning human processing of faces that have been drawn by the psychology researchers and neuroscientists are described. Then the most important tasks related to the computer facial processing are shown: face detection, face recognition and processing of facial features, and the main strategies as well as the methods applied in the related fields are presented. Finally, the applications of digital biometrical processing of human faces are presented.
Chapter Preview
Top

Introduction

The recognition of faces is one of the easiest and the most frequently used by adult human method of distinction between known and unknown persons and identification of these familiar. Usually we benefit from it even without the awareness of this process and similarly another people can recognize our identity basing only on the quick observation. Often such recognition is done even without our knowledge or our permission to do it, and moreover, usually we ignore the most of looks given us. It is not meaningless, that recognition of faces, along with the recognition of voice, is the most common in the nature method of identification of individuals for a number of species.

Automatic processing of faces has found a number of applications in several specific fields: in security, forensic and commercial solutions. The neutrality and non-intrusiveness are the main reasons why the automatic machine face recognition systems are often treated as the promising tool for security applications. The precision of facial recognition by humans motivates researchers to apply the artificial facial recognition in forensics. Moreover, recognition of faces is very natural to all human beings, and it does not arouse such negative (criminal-like) associations as, for example, gathering fingerprints or samples of the genetic material for DNA tests, thus biometrical processing of faces can be successfully used in commercial applications.

Faces and their recognition always have been of the interest to various researchers. However, the dynamic growth of interest in this field we can observe from early 1990s. There are several reasons that motivate the progress of research in the automatic facial recognition. One of the most important is the development of the hardware, which has allowed the real-time acquiring and processing of vision data. Also the rising both importance and number of observation surveillance systems cause the interest in such biometric technologies. It is not meaningless, that along with developing the knowledge on biometrical recognition of faces to the mature level, it becomes the regular product available on the commercial market.

The automatic face processing is a field, which attract the attention of researchers in various disciplines of engineering, such as signal and image processing, pattern recognition, machine vision, computer graphics. Face perception by humans is also interesting for scientists, mainly for psychologists and neuroscientists. What is important, the results of observations concerning face recognition in humans can pose an inspiration for engineering research on machine methods, and also contrary to that, results of research on automatic face processing may suggest interesting directions of research in humans.

Automatic face processing is a non-trivial task, and the facial researches, independent on the final application of the issue, need to face a series of challenges, mainly derived from the fact, that machine processing of faces from images (or video sequences) concerns the classification of three-dimensional elastic objects, while the data are available merely in a two-dimensional form. When face recognition is evaluated in controlled conditions it shows very good performance, but when it is applied to real-world applications, especially in non-controlled conditions, it still does not achieve such good results as recognition of fingerprint or iris. Nevertheless, these better performing biometrics definitely require to cooperate (thus also the goodwill) from the examined subject, e.g. to contact with fingerprint scanner or to present the eye in location proper to iris sensor.

Key Terms in this Chapter

Face Authentication: A face processing task that concerns the one-to-one matching for verification or rejection of the assumed identity of photographed person. This task may be also labeled as the face verification

LDA: LDA stands for Linear Discrimant Analysis. Also known as Fisher Discriminant Analysis (FDA) or Fisher Linear Discrimant (FLD). This is popular in face recognition method of dimensionality reduction that search for vectors that best discriminate among classes.

Eigenfaces: Face-like two-dimensional representation of basis vectors obtained by eigen-decomposition (PCA) of set of facial images.

Face Expression Analysis: A face processing problem concerning the recognition of basic affective states from facial photographs. This should not be mistaken with emotion recognition, which is task based on higher level knowledge.

Face Recognition: A face processing task, which goal is to identify the identity of the person put in the image. This relates to the one-to-many matching. This task may be also termed as face identification.

Face Tracking: A face processing task that concerns continuous detection and localization of face (or faces) in the video sequence, and usually should be able to perform these operations in the real time.

Fisherfaces: Face-like two-dimensional representation of basis vectors obtained by Linear (Fisher’s) Discriminant Analysis on set of facial images.

PCA: PCA stands for Principal Component Analysis. Also known as Karhunen-Loeve Transform (KLT). This is popular in face recognition method of dimensionality reduction of data sets, based on eigen-decomposition of covariance matrix.

Face Detection: A face processing problem related to which aims to determine whether or not any human faces are present in the given input image, and if they are return the information about location and extent of each face in the image.

Complete Chapter List

Search this Book:
Reset