IEEE TRANSACTIONS ON IMAGE PROCESSING, VOL. 15, NO. 11, NOVEMBER 2006 3349 Face Verification Across Age Progression Narayanan Ramanathan, Student Member, IEEE, and Rama Chellappa, Fellow, IEEE Abstract—Human faces undergo considerable amounts of varia- tions with aging. While face recognition systems have been proven to be sensitive to factors such as illumination and pose, their sen- sitivity to facial aging effects is yet to be studied. How does age progression affect the similarity between a pair of face images of an individual? What is the confidence associated with establishing the identity between a pair of age separated face images? In this paper, we develop a Bayesian age difference classifier that classifies face images of individuals based on age differences and performs face verification across age progression. Further, we study the sim- ilarity of faces across age progression. Since age separated face im- ages invariably differ in illumination and pose, we propose prepro- cessing methods for minimizing such variations. Experimental re- sults using a database comprising of pairs of face images that were retrieved from the passports of 465 individuals are presented. The verification system for faces separated by as many as nine years, attains an equal error rate of 8.5%. Index Terms—Age progression, face recognition, face verifica- tion, probabilistic eigenspaces, similarity measure. I. INTRODUCTION P ERCEIVING human faces and modeling the distinctive features of human faces that contribute most towards face recognition are some of the challenges faced by computer vision and psychophysics researchers. Human faces belong to a spe- cial class of 3-D objects, modeling which involves developing accurate characterizations that account for illumination, head pose variations, facial expressions, etc. Moreover, human faces also undergo growth related changes that are manifested in the form of shape and textural variations. Hence, the robustness to variations due to factors such as illumination, pose, facial expressions, aging, etc., is a significant metric in evaluating face recognition systems. Over the years, many still-image and video based face recognition algorithms have been developed. Zhao et al. [2] provide a thorough qualitative analysis of the many different face recognition algorithms. Recently, good to very good recognition performance across illumination and pose variations were demonstrated by Zhang and Samaras [3] and Zhou and Chellappa [4], Blanz and Vetter [5], Georghiades Manuscript received October 14, 2005; revised April 17, 2006. A portion of the work presented in this paper was presented at the IEEE Conference on Com- puter Vision and Pattern Recognition. This work was supported in part by a fel- lowship from P. Horvitz (Apptis, Inc.). The associate editor coordinating the review of this manuscript and approving it for publication was Dr. Ercan E. Ku- ruoglu. N. Ramanathan is with the Department of Electrical and Computer Engi- neering, University of Maryland, College Park, MD 20742-3275 USA (e-mail: ramanath@umiacs.umd.edu). R. Chellappa is with the Center for Automation Research (CfAR) and the Department of Electrical and Computer Engineering, University of Maryland, College Park, MD 20742-3275 USA (e-mail: rama@umiacs.umd.edu). Color versions of Figs. 5, 8, and 9 are available online at http://ieeexplore. ieee.org. Digital Object Identifier 10.1109/TIP.2006.881993 et al. [6], and Gross et al. [7]. Moreover, many approaches have been proposed for modeling facial expressions. Facial Action Coding systems [8] have contributed significantly towards char- acterizing facial expressions. Yacoob and Davis [9], Essa and Pentland [10], Martinez [11], Tian et al. [12], and Liu et al. [13] have proposed approaches for identifying facial expressions from face images. Decades of dedicated research coupled with the advent of standardized performance evaluation protocols such as FERET [14], [15], and FRVT [16] have enhanced the commercial significance of face recognition systems. Though psychophysical studies have contributed signifi- cantly towards the perception of growing faces and towards understanding craniofacial growth, age progression in human faces has largely been ignored while developing face recogni- tion systems. Modeling age progression in human faces is a very challenging task. Facial aging effects are manifested in different forms in different age groups. While aging effects are often manifested in the form of shape variations in human faces due to the cranium’s growth from infancy to teen years [17], they are more commonly observed in the form of textural variations such as wrinkles and other skin artifacts in adult faces. Apart from biological factors, factors such as climatic conditions, ethnicity, mental stress, etc., are often attributed to play a role in the process of aging. Some of the interesting applications of studying age progression in human faces are discussed as follows. • Developing face recognition systems that are robust to age progression would enable the successful deployment of face recognition systems in public places. Such systems would be highly beneficial to homeland security applications. Fur- ther, developing systems that verify face images across age progression would annul the necessity of periodically up- dating large face databases with more recent images. • Since different individuals age differently, developing automatic age progression systems that could predict the many different ways a person could have aged would have a significant impact in finding missing individuals. • Ethological studies have revealed that the perceived age of an individual significantly affects the type and amount of be- havior directed towards him/her by other individuals. Hence, building systems that could reliably estimate the age of indi- viduals, would be useful for developing human-robot inter- action systems and human-computer interaction systems. • Changes in facial appearances are attributed to have a sig- nificant psychosocial impact on an individual [17]. Studies related to craniofacial growth are bound to help surgeons and orthodontists in treating disfigurements and deformi- ties in faces. Before we formulate the problem, we provide a brief overview of the previous work on age progression in human faces. 1057-7149/$20.00 © 2006 IEEE