Recognizing Face Images with Disguise Variations

Recognizing Face Images with Disguise Variations

Neslihan Kose (EURECOM, France), Jean-Luc Dugelay (EURECOM, France), Richa Singh (IIIT – Delhi, India) and Mayank Vatsa (IIIT – Delhi, India)
Copyright: © 2014 |Pages: 25
DOI: 10.4018/978-1-4666-5966-7.ch011
OnDemand PDF Download:
$30.00
List Price: $37.50

Abstract

Challenges in automatic face recognition can be classified in several categories such as illumination, image quality, expression, pose, aging, and disguise. In this chapter, the authors focus on recognizing face images with disguise variations. Even though face recognition with disguise variations is a major challenge, the research studies on this topic are limited. In this study, first disguise variations are defined followed by an overview of the existing databases used for disguise analysis. Next, the studies that are dedicated to the impact of disguise variations on existing face recognition techniques are introduced. Finally, a collection of several techniques proposed in state-of-the-art which are robust against disguise variations is provided. This study shows that disguise variations have a significant impact on face recognition; hence, more robust approaches are required to address this important challenge.
Chapter Preview
Top

1. Introduction

Face recognition is used with diverse security contest such as surveillance, border security and forensic investigation. Face recognition is a process in which an individual is identified or verified based on facial characteristics. There have been several researches for face recognition in the presence of pose, expression, and illumination variations (Zhao et al., 2003;Li & Jain, 2005;Wechsler, 2006;Delac & Grgic, 2007). In recent face recognition studies, the results show that under normal changes in constrained environment, the performance of existing face recognition systems is greatly enhanced. However, in most real world applications, there can be several problems such as using images of low quality or non-cooperative users or temporal variations and dissimilarities in facial characteristics that are created using disguise accessories.

Challenges in face recognition can be listed as illumination, image quality, expression, pose, aging, and disguise. Among these challenges, recognition of faces with disguise is a major challenge and has been recently addressed by several researchers (Alexander & Smith, 2003;Ramanathan et al., 2004;Silva & Rosa, 2003;Singh et al., 2008). Disguise accessories can be used to alter the appearance of an individual, to impersonate another person, or to hide one’s identity. In the book chapter (Singh et al., 2008), some studies on disguise variations are explained. According to this chapter, (Ramanathan et al., 2004) studied facial similarity for several variations including disguise by forming two eigenspaces from two halves of the face, one using the left half and other using the right half. From the test image, optimally illuminated half face is chosen and is projected into the eigenspace. This algorithm has been tested on the AR face database (Martinez & Benavente, 1998) and the National Geographic database (Ramanathan et al., 2004) which consists of variations in smile, glasses, and illumination. An accuracy of around 39% for best two matches is reported on the AR database. Silva and Rosa (2003) uses Eigen-eyes to handle several challenges of face recognition including disguise. Using the Yale face database (Belhumer et al., 1997), the algorithm was able to achieve an accuracy of around 87.5%. The advantage of the algorithm is that alterations in facial features excluding the eye region do not affect the accuracy. Pamudurthy et al. (2005) proposed a face recognition algorithm which uses dynamic features obtained from skin correlation and the features are matched using nearest neighbor classifier. On a database of 10 individuals, authors reported that this approach gives accurate results. Alexander and Smith (2005) used PCA based algorithm with Mahalanobis angle as the distance metric. The results show an accuracy of 45.8% on the AR database. The limitation of these algorithms is that the performance degrades when important regions such as the eye and mouth are covered. Moreover, in Singh et al. (2008), it is stated that the AR and Yale databases do not contain many images with disguise and therefore are not ideal for validating algorithms under comprehensive disguise scenarios.

Complete Chapter List

Search this Book:
Reset