PERFORMANCE EVALUATION OF NON-IDEAL IRIS BASED RECOGNITION SYSTEM IMPLEMENTING GLOBAL ICA ENCODING Vivekanand Dorairaj, Natalia A. Schmid, and Gamal Fahmy * * This work was supported by a grant from NSF IUCRC Center for Identification Technology Research. Lane Department of Computer Science and Electrical Engineering West Virginia University, P.O. Box 6109, Morgantown, WV 26506 {vivekand,natalias,fahmy}@csee.wvu.edu ABSTRACT We describe and analyze the performance of a non-ideal iris recognition system. The system is designed to process non-ideal iris images in two steps: (i) estimation of the gaze direction and (ii) processing and encoding of the rotated iris image. We use two objective functions to estimate the gaze direction: Hamming distance and Daugman’s integro-differential operator and determine an estimated angle by picking the value that optimizes the selected objective function. After the angle is estimated, the off-angle iris image undergoes geometric transformations involving the estimated angle and is further processed as if it were a frontal view image. The encoding technique developed in this work is based on application of the global Independent Component Analysis (ICA) to masked iris images. We use two datasets: CASIA dataset and a special dataset of off- angle iris images collected at WVU to verify the performance of the encoding technique and angle estimator, respectively. A series of Receiver Operating Characteristics (ROCs) demonstrates various effects on the performance of the non-ideal iris based recognition system implementing the global ICA encoding. 1. SYSTEM DESCRIPTION Iris patterns are believed to be unique due to the complexity of two underlying processes (i) environmental and (ii) genetic that influence their generation. These result in textural patterns that are unique to each eye of an individual and even distinct between twins. Iris as a biometric has been known for a long time [1-4]. However, only over the past two years it has gained a substantial attention of both the research community and governmental organizations. Three critical factors that influenced the increased interest to iris biometric are (i) public acceptance, (ii) new user friendly capture devices with broad improved capabilities, and (iii) a broadened range of applications. As a result, a large number of new iris encoding and processing techniques have been developed over this short period of time. While most of literature is focused on processing of frontal view iris images [2,3,4], there have been a few new directions identified in iris biometric research including processing and encoding of “non-ideal iris” that is defined as dealing with off-angle, occluded, blurred, noisy images [8,9,10] and “iris at a distance” identified as a video or a snapshot of iris captured from a not necessarily cooperative individual at a large distance (more than a meter) [11]. In this work, we design a non-ideal iris recognition system that deals with off-angle iris images, and analyze its performance. The system processes non-ideal iris images in two steps: (i) estimation of a gaze direction and application of a projective transformation to bring the iris image into a frontal view image and (ii) processing and encoding of the rotated iris image as if it were a frontal view image. To estimate the gaze direction we use the Hamming distance between an ideal frontal view image and an off-angle iris image or Daugman’s integro-differential operator [1]. A brief description of the angle estimation strategy is given in Sec. 2. The iris image is further enhanced, segmented using the integro-differential operator, and transformed into a pseudo-polar representation (see Fig.1). While a set of standard preprocessing steps similar to those described in [1-4] is used to prepare iris image for encoding, the encoding technique introduced and evaluated in this work is quite distinct from all previous techniques. We use the global ICA method for encoding the iris texture. We are aware of a few previously published works that use ICA method for iris image encoding (for example, [6]). However, in all these works the ICA was used in a mode of operation that extracts only local features, as proposed by Hyvarinen [7]. The purpose of this research is to explore a possibility of using global image encoding / feature extraction algorithms to process the iris. Apart from this, we extract individual iris signatures and demonstrate their independence, which results in a simplified predictive analysis of iris individuality (not presented in this paper). Prior to extracting ICA components, we perform PCA that is often used as a preprocessing step to ICA with the goal to uncorrelate components [7]. 0-7803-9134-9/05/$20.00 ©2005 IEEE