A reliable iris recognition algorithm based on reverse biorthogonal wavelet transform R. Szewczyk ⇑ , K. Grabowski, M. Napieralska, W. Sankowski, M. Zubert, A. Napieralski Department of Microelectronics and Computer Science, Technical University of Lodz, Wolczanska Street 221/223 Building B18, 90-924 Lodz, Poland article info Article history: Available online 7 September 2011 Keywords: Biometrics Iris recognition Signatures encoding Signatures matching Reverse biorthogonal wavelet transform abstract This article describes an iris recognition algorithm designed to analyze noisy iris biometric data. The methods forming part of the authentication process were developed and optimized by the authors using visible wavelength images of an eye taken under unconstrained conditions (at a different perspectives, illuminations, occlusion grades, etc.), mainly contained in the UBIRIS.v2 database. The whole iris authen- tication system was submitted by the authors to the International Iris Recognition Contest NICE.II, where it took eighth place, while the iris segmentation stage itself took second place in the previous contest — NICE.I. This paper is focused on the iris feature extraction stage — the method developed by the authors to ana- lyze noisy iris biometric data. Several techniques used for more efficient and robust analysis of such images and issues concerning the best wavelet selection are also presented in this paper. Ó 2011 Elsevier B.V. All rights reserved. 1. Introduction One of the open problems in biometrics that limits the applicabil- ity of iris biometrics is iris identification at-a-distance and on-the- move that must be done in unconstrained imaging conditions and using potentially large databases. To broaden the usability of iris biometrics, additional analysis techniques must be developed, tak- ing into account the specificity of the degraded images, such as light reflections from the eye’s surface, occlusions and fluctuations of per- spective and illumination. Most of these factors limit availability of distinctive features needed for proper recognition. Thus, recently, significant effort has been focused on authenticating objects at-a- distance and on-the-move using the iris biometrics (Proenca and Alexandre, 2010; Matey et al., 2006). This paper describes a recogni- tion strategy, which can be used for such purposes. The presented solution was tested within the Noisy Iris Challenge Evaluation — Part II (NICE:II) contest using the UBIRIS.v2 database (Proenca et al., 2010). In the contest, the proposed solution took eighth place. 2. Description of the proposed method In general, the proposed strategy, presented as a flowchart in Fig. 1, is similar to a well-known scenario of the authentication process. The identification starts when an image of the eye from a person located in front of the digital camera is acquired. In the pro- cess of image acquisition, a digital representation of the biometrics T x is obtained from a real biometrics T x . The next step is iris image segmentation, which identifies the area of interest, the iris texture, and transforms it into the pseudo-polar coordinate system (Daug- man, 1993). In the following step, the segmented iris image is addi- tionally processed using blue channel removal, image conversion to monochromatic, eyelid occlusions and reflections removal, eye- lashes removal and resulting iris image histogram equalization. The extraction of distinctive features and the encoding phase is a process that allows a digital template b T x of the biometrics to be obtained. The resulting template, which is 324 bits wide and is ob- tained using reverse biorthogonal 3.1 wavelet, is compared to those in a database using similarity score s to search for the pattern that best matches the template using a decision threshold t. This threshold is chosen based on the assumed security strategy — see (Bolle et al., 2004). Each part of the algorithm is explained in detail in consecutive subsections below. 2.1. Image segmentation The image segmentation algorithm, described in detail else- where (Sankowski et al., 2010; Sankowski, 2009), localizes an iris in the eye image and transforms the iris region to the pseudo-polar coordinate system (Daugman, 1993) to obtain a rectangular image of iris structure, as presented in Fig. 2. The segmentation algorithm proposed by the authors was submitted to the NICE.I contest (Proenca and Alexandre, 2010), where it took second place. 0167-8655/$ - see front matter Ó 2011 Elsevier B.V. All rights reserved. doi:10.1016/j.patrec.2011.08.018 ⇑ Corresponding author. E-mail addresses: szewczyk@dmcs.p.lodz.pl (R. Szewczyk), kgrabowski@dmcs. p.lodz.pl (K. Grabowski), mnapier@dmcs.p.lodz.pl (M. Napieralska), wsan@dmcs.p.lodz.pl (W. Sankowski), mariuszz@dmcs.p.lodz.pl (M. Zubert), napier@dmcs.p.lodz.pl (A. Napier alski). URL: http://www.irisep.dmcs.pl (R. Szewczyk). Pattern Recognition Letters 33 (2012) 1019–1026 Contents lists available at SciVerse ScienceDirect Pattern Recognition Letters journal homepage: www.elsevier.com/locate/patrec