Beyond One Still Image Face Recognition from Multiple Still Images or Video Sequence.pdf

Beyond One Still Image: Face Recognition from Multiple Still Images or Video Sequence Shaohua Kevin Zhou 1 and Rama Chellappa 2 1 Integrated Data Systems Department Siemens Corporate Research, Princeton, NJ 08540 Email: {kzhou}@ 2 Center for Automation Research and Department of Electrical and Computer Engineering University of Maryland, College Park, MD 20742 Email: {rama}@ Abstract While face recognition from a single still image has been extensively studied over a decade, face recognition based on a group of still images (also referred as multiple still images) or a video sequence is an emerging topic. Although a group of still images or a video sequence can be treated as a single still image in a degenerate manner, additional properties are present in multiple still images and/or video sequences. In this chapter, we summarize three properties manifested in multiple still images and/or video sequence. Then we review recently proposed approaches utilizing the manifested properties. Finally, we forecast feasible approaches that could realize the full potentials of multiple still images or video sequences. Index Terms Face recognition from one still image, face recognition from multiple still images, face recognition from video, statistical analysis. I. INTRODUCTION While face recognition (FR) from a single still image has been extensively studied over a decade, FR based on a group of still images (also referred to as multiple still images) or a video sequence is an emerging topic. This is mainly evidenced by the growing increase in the literature. For instance, a research initiative called Face Recognition Grand Challenge [1] has been organized. One specific challenge directly addresses the use of multiple still images that is reportedly to improve the recognition accuracy significantly [3]. Recently a workshop jointly held with CVPR 2004 was devoted to face processing in video [2]. It is predictable that with the ubiquity of video sequences, face recognition based