The paper reviews face recognition techniques—an actively researched area in the field of biometrics, pattern recognition and computer vision. Maiden attempts were made in the early 1960s or so, but significant progresses were made only around 1988, in synchronization with a massive increase in computational power. The first widely accepted algorithm was the Principal Component Analysis (PCA) or Eigenface method, which even today is used as the most significant tool for dimensionality reduction. Today, many scientists agree that the case of any two simple facial images, under well-controlled conditions of environmental constraints/variables, for comparison is practically known to be a solved problem. Even with minimal variations in such images, apart from facial expression, the problem is insignificant by today’s standards with a recognition accuracy of around 90-98% reported across many papers. This is arguably even better than human performance in the same conditions provided, when humans are tested on the images of the unknown persons. However, when variations in images caused by pose, aging or extreme illumination and pattern recognition conditions are introduced, the ability of human beings to recognize faces is still remarkable as compared to computers.

Nowadays, the use of image analysis and its applications in facial recognition and other biometric-based systems has been on the increase very rapidly with possible utilizations obviously in the fields of identity verification of individuals, secured and privileged access control systems, law enforcement, e-banking, entertainment, smart homes, and numerous other domains. Based on the face representation used for the recognition, face recognition techniques can be further subdivided into two main groups. First is appearance-based and the second is feature-based. Appearance-based methods consider the global properties of the face and use the whole face image (or some specific image regions) to extract facial features, while feature-based methods, in contrast, are based on local facial characteristics (such as eyes, nose and mouth) and use parameters such as angles and distances between ducial points on the face as descriptors for face recognition. Human physiological or behavioral characteristics are available uniquely to each individual as biometric evidences. Unimodal biometric systems are designed based on a single identifier, such as face or a fingerprint, etc. However, it suffers from several limitations, and due to these limitations, it often degrades the performance. These limitations can be reduced by means of multimodal biometric systems that combine the voids created by one over the other evidences once obtained by multimode or multiple sources used at image capturing stage only.

Telecommunications Journal, Face Recognition Technique (FRT), Principal Component Analysis (PCA), Linear Discriminant Analysis (LDA), Eigenvalue, Eigenvector, Eigenspace, Eigenface.