NIR and VW iris image recognition using ensemble of patch statistics features

• Published in: The Visual computer Vol. 35; no. 9; pp. 1327 - 1344
• Main Authors: Umer, Saiyed, Dhara, Bibhas Chandra, Chanda, Bhabatosh
• Format: Journal Article
• Language: English
• Published: Berlin/Heidelberg Springer Berlin Heidelberg 01.09.2019; Springer Nature B.V
• Subjects: Artificial Intelligence; Biometrics; Classification; Computer Graphics; Computer Science; Feature extraction; Fourier transforms; Image acquisition; Image Processing and Computer Vision; Image segmentation; Infrared imagery; Localization; Methods; Neural networks; Original Article; Parameter estimation; Preprocessing; Visible wavelength; Ensemble of patch statistic; Identification; Near-infrared; Iris recognition
• ISSN: 0178-2789; 1432-2315
• DOI: 10.1007/s00371-018-1544-4

A novel iris recognition system is proposed in this paper. The proposed system is able to handle various challenging issues which may occur during image acquisition in near-infrared and/or visible wavelength lights under constrained and less-constrained environments. The proposed system demonstrates great perseverance for recognizing subjects in both stable and adverse situations. During recognition, the system performs image preprocessing, feature extraction, and classification tasks. During preprocessing, an annular iris portion is segmented out from an input eyeball image, and for this, two different segmentation approaches: one for near-infrared images and another for visible wavelength images, have been proposed. A novel patch-based histogram-type feature (ensemble of patch statistics) which adopts a statistical approach of texture analysis is employed during feature extraction. For the proposed system, the extensive experimental results have been demonstrated using ten benchmark iris databases, namely MMU1, UPOL, IITD, UBIRIS.v1, CASIA-Interval-v3, CASIA-Iris-Twins, CASIA-Iris-Thousand, CASIA-Iris-Distance, CASIA-Iris-Syn, and UBIRIS.v2. The performance of the proposed system is compared with the state-of-the-art methods on these databases and the comparisons show significant out-performance on the competing methods.