Chapter 17 - Beyond one still image: Face recognition from multiple still images or a video sequence

Abstract

This chapter reviews the emerging literature on FR using more than one still image—that is, multiple still images or a video sequence. While face recognition (FR) from a single still image has been extensively studied over a decade, FR based on a group of still images or a video sequence is an emerging topic. This is proved by the growing increase in the literature. For example, a research initiative called Face Recognition Grand Challenge is organized that addresses the use of multiple still images to improve the recognition accuracy significantly. Multiple still images or a video sequence can be regarded as a single still image. Studying the recognition algorithms from the perspective of additional properties is very beneficial in many ways: easy categorization of the algorithms available in the literature accordingly and development of new approaches to realize the full potentials of multiple still images or video sequences. The chapter details the basics of face recognition, its properties, and reviews FR approaches utilizing the three properties—namely, multiple observations, temporal continuity/dynamics, and 3D model. A graph comparing the conventional subspace method (CSM) and mutual subspace method (MSM) is also presented. In a degenerate manner, FR from more than one image can be reduced to recognition based on a single still image. Such a treatment ignores the properties additionally possessed by more than one image. The chapter also suggests potential approaches that utilizes the full benefits of multiple still images or video sequences.