IEEE TRANSACTIONS ON SIGNAL PROCESSING, VOL. 46, NO. 4, APRIL 1998 1185 A Human Identification Technique Using Images of the Iris and Wavelet Transform W. W. Boles and B. Boashash Abstract—A new approach for recognizing the iris of the human eye is presented. Zero-crossings of the wavelet transform at various resolution levels are calculated over concentric circles on the iris, and the resulting one-dimensional (1-D) signals are compared with model features using different dissimilarity functions. Index Terms— Human identification, iris recognition, wavelet trans- form. I. INTRODUCTION Computer vision-based techniques that recognize human features such as faces, finger prints, palms, and eyes have many applications in surveillance and security. Most of the existing methods have limited capabilities in recognizing relatively complex features in realistic practical situations. The objective of this correspondence is to present a new approach for recognizing humans from images of the iris of the eyes under practical conditions. The iris has unique features and is complex enough to be used as a biometric signature [1]. This means that the probability of finding two people with identical iris patterns is almost zero. Therefore, in order to use the iris pattern for identification, it is important to define a representation that is well adapted for extracting the iris information content from images of the human eye. We propose a new algorithm for extracting unique features from images of the iris of the human eye and representing these features using the wavelet transform (WT) zero crossings [2]. This representation is then utilized to recognize individuals from images of the irises of their eyes. A wavelet function that is the first derivative of a cubic spline is used to construct the representation. The proposed technique is translation, rotation, and scale invariant. It is also largely unaffected by variations in illumination and noise levels in the images. II. REPRESENTATION OF IRIS PATTERNS The proposed iris recognition system is designed to handle noisy conditions as well as possible variations in illumination and camera- to-face distances. In studying the characteristics of the irises, we will only deal with samples of the grey-level profiles and use these to construct a representation. Input images are preprocessed to extract the portion containing the iris. We then proceed to extract a set of one dimensional (1-D) signals and obtain the zero-crossing representations of these signals. The main idea of the proposed technique is to represent the features of the iris by fine-to-coarse approximations at different resolution levels based on the WT zero- crossing representation. To build the representation, a set of sampled data is collected, followed by constructing the zero-crossing repre- sentation based on its dyadic WT. Manuscript received February 15, 1997; revised November 30, 1997. The associate editor coordinating the review of this paper and approving it for publication was Dr. Ahmed Tewfik. The authors are with the Signal Processing Research Centre, School of Electrical and Electronic Syetems Engineering, Queensland Univer- sity of Technology, Brisbane, Australia (e-mail: w.boles@qut.edu.au; b.boashash@qut.edu.au). Publisher Item Identifier S 1053-587X(98)02926-2. Fig. 1. Sample image. The process of information extraction starts by locating the pupil of the eye, which can be done using any edge detection technique. Knowing that it has a circular shape, the edges defining it are connected to form a closed contour. The centroid of the detected pupil is chosen as the reference point for extracting the features of the iris. The grey level values on the contours of virtual concentric circles, which are centered at the centroid of the pupil, are recorded and stored in circular buffers. In what follows, for simplicity, one such data set will be used to explain the process and will be referred to as the iris signature. Choosing the centroid as the reference point ensures that the representation is translation invariant. We now need to compensate for size variations due to the possible changes in the camera-to-face distance. The extracted data from the same iris may be different even if the diameter of the used virtual circle is kept constant. This is due to the possible variation in the size of the iris in the image as a result of a change in the camera-to-face distance. For matching purposes, the extracted data must be processed a) to ensure accurate location of the used virtual circle and b) to fix the sample length before constructing the zero-crossing representation. Using the edge-detected image, the maximum diameter of the iris in any image is calculated. In comparing two images, one will be considered to be a reference image. The ratio of the maximum diameter of the iris in this image to that of the other image is also calculated. This ratio is then used to make the virtual circles, which extract the iris features, have the same diameter. In other words, the dimensions of the irises in the images will be scaled to have the same constant diameter regardless of the original size in the images. Furthermore, the extracted information from any of the virtual circles must be normalized to have the same number of data points. We introduce a normalization value , which is selected as a power-of-two integer. The main reason for this selection is to enable the extraction of the whole information available in the iris signature by applying the dyadic wavelet transform. By changing the normalization constant, the accuracy of the classification process can be adjusted. A large value of results in decomposing the iris signature to a large number of levels, in which the information of iris signature is analyzed in more detail. This implies that the classification is more accurate. In contrast, a small normalization value results in reducing the accuracy of the classification but increases the speed of the whole process. Fig. 1 shows a sample image. The extracted data set and its wavelet transform are shown in Fig. 2. The next step is to generate a zero-crossing representation from the normalized iris signature . Since the normalized iris signature represents a closed ring, it is naturally periodic with period , and the zero-crossing representation will also be periodic since the wavelet coefficients are periodic. This means that the representation 1053–587X/98$10.00  1998 IEEE