Iris localization using local histogram and other image statistics Muhammad Talal Ibrahim a,n , Tariq M. Khan b , Shahid A. Khan b , M. Aurangzeb Khan b , Ling Guan a a Ryerson Multimedia Research Lab, Ryerson University, Toronto, Canada b Department of Electrical Engineering, COMSATS Institute of Information Technology, Islamabad, Pakistan article info Article history: Received 16 April 2011 Received in revised form 21 September 2011 Accepted 13 November 2011 Available online 13 January 2012 Keywords: Iris localization Iris recognition Local histogram Standard deviation and gradient abstract This paper presents an automatic method for iris localization based on image statistics. The proposed method localizes the iris in two stages. In the first stage, a circular moving window is used to localize the pupil by finding the range of grey levels that has the highest probability of enclosing the pupil. The window with the grey level peak having the minimum standard deviation of x- and y-coordinates is selected as the region enclosing the pupil. In the second stage, effect of the eyelashes is reduced by using median filtering and the iris boundary is estimated by taking the gradient of the rows within pupil. The proposed method has been tested on three public databases: CASIA-IrisV1, CASIA-IrisV3- Lamp and MMU version 1.0. Experimental results demonstrate the superiority of the proposed method in comparison with some of the existing methods. & 2011 Elsevier Ltd. All rights reserved. 1. Introduction In the last two decades, biometric based security systems have gained popularity. A number of biometric techniques have been developed and introduced for person identification and verifica- tion. In the literature, there are as many biometric techniques available as there are many unique/identifiable features in a person [1]. Generally, biometric deals with recognizing the humans based on their physical or behavioral traits. Physical traits include face, fingerprint, iris, palm print, hand geometry, and ear shape. Physical traits are considered to be unchangeable, whereas behavioral traits like hand written signatures, gait, key- strokes, and voice are the reflection of human behavior and they tend to vary with time [1]. Among all of the mentioned biometric traits, iris is considered to be the most reliable, stable, and unique biometric trait [2]. The iris is an area of the eye where the pigmented or the colored circle rings the dark pupil as shown in Fig. 1. Generally, iris is considered as a circular shaped region between the pupil and the sclera. It is an internal organ that can be seen outside the body and is well protected from the environment [3]. It consists of number of features like freckles, coronas, stripes, arching, ligaments, ridges, crypts, rings, furrows, and zigzag collarette [2,4–6]. These features are statistically unique and stable [7]. Furthermore, the distribution of these features in the human iris is random [8]. The human iris tends to change very little throughout the individual’s life. Due to these properties, the iris is considered as one of the most secure and reliable characteristic of a human that is most suitable for person identification [9]. It is also observed that not only the irises of two twins are different, but also the irises of one person are not identical [2]. Generally, the performance of an iris recognition system highly depends on iris localization and segmentation. Iris segmentation deals with isolation of the iris from other parts of an eye like pupil, sclera, surrounding skin, eyelashes, and eyebrows. It has been observed that the most computationally intensive task in iris recognition, especially in iris segmentation is to accurately determine the inner and outer boundaries of iris. Generally, the methods proposed for iris localization can be divided into two major categories. The first category is based on circular edge detectors like Hough transform, and the second category is based on the histogram. In this paper, a local histogram and standard deviation based iris localization method has been proposed. The proposed method starts by finding the range of grey levels that has the highest probability of enclosing the pupil by using a circular moving window. Once the pupil is localized, the effect of scattered eyelashes is minimized and the boundary of the iris is extracted based on the peaks of the first derivative of the rows within the pupil. Rest of the paper is organized as follows. Section 2 covers the background and the related work. Section 3 deals with the design and structure of the proposed method. Section 4 presents the experimental results. Section 5 explains the limitations of the proposed system. Finally, conclusions are presented in the last section of this paper. Contents lists available at SciVerse ScienceDirect journal homepage: www.elsevier.com/locate/optlaseng Optics and Lasers in Engineering 0143-8166/$ - see front matter & 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.optlaseng.2011.11.008 n Corresponding author. E-mail addresses: muhammadtalal.ibrahi@ryerson.ca (M.T. Ibrahim), tariq_mehmood@comsats.edu.pk (T.M. Khan), shahidk@comsats.edu.pk (S.A. Khan), aurangzeb_niazi@comsats.edu.pk (M.A. Khan), lguan@ee.ryerson.ca (L. Guan). Optics and Lasers in Engineering 50 (2012) 645–654