Iris Recognition Using Level Set and Local Binary Pattern

Abstract

Abstract-This paper presents an efficient algorithm for iris recognition using the Level Set (LS) method and Local Binary Pattern (LBP). We deploy a Distance Regularized Level Set (DRLS)-based iris segmentation procedure in which the regularity of the Level Set (LS) function is intrinsically maintained during the curve propagation process. The LS evolution is derived as the gradient flow that minimizes energy functional with a distance regularization term and an external energy that drives the motion of the zero LS toward iris boundary accurately. DRLS also uses relatively large time steps in the finite difference scheme to reduce the curve propagation time. The deployed variational model is robust against poor localization and weak iris/sclera boundaries. Furthermore, we apply a Modified LBP (MLBP) in an effort to elicit the iris feature elements. The MLBP combines both the sign and magnitude features for the improvement of iris texture classification performance. The identification and verification performance of the proposed scheme is validated using the CASIA version 3 interval dataset.Index Terms-Iris recognition, distance regularized level set, modified local binary pattern. I. INTRODUCTIONMost current iris recognition algorithms perform relatively well in a strictly controlled environment. However, the iris images captured in a noisy environment produce nonideal iris images with a varying image quality and are severely affected from eyelid and eyelash occlusions, motion blurs, camera diffusions, head rotations, gaze directions, camera angles, and reflections [1]. Furthermore, the iris contours of the noisy irises are not exactly circular or elliptical and may be of any kind of shapes [2]- [6]. To mitigate the shape irregularities of the iris contours, several researchers proposed different iris segmentation methods based on active contours, including the modified Mumford-Shah segmentation model [2], Variational Level Set (VLS) [3], Fourier series expansions of the contour data [4], and Geodesic Active Contours (GAC) [5].The segmentation approaches based on the traditional active contours may involve a huge computational time. Also, the parametric active contour-based iris segmentation scheme may be terminated due to specular reflections, the Manuscript received October 30, 2013; revised February 17, 2014 thick radial fibres in the iris and the crypts in the ciliary region. In conventional active contour method with the LS formulation, the LS function typically develops irregularities during its evolution, which may cause numerical errors and eventually destroy the stability of the evolution [7]. Addressing the above problems, we apply the Distance Regularized Level Set (DRLS) method, proposed in [7], to localize the iris contour. In the DRLS method, the regularity of the LS function is intrinsically maintained during the curve propagation. The LS evolution is derived as the gradient flow that minimizes energy functional with a distance regularization term and an external energy that drives the motion of t...

Feature extraction using sequential cumulative bin and overlap mean intensity for iris classification

“…Finding local attributes for iris descriptors is common practice at the feature extraction level in iris recognition [17][18][19]. Local binary pattern (LBP) is the prevalent approach that uses local information of images for features, and it has been widely used in various application areas including face, texture, and iris * Correspondence: nazriali@usm.my recognition as well [20][21][22][23]. First introduced by Ojala et al in 1996, this basic LBP operator was generalized to multiscale and multisampling variants of LBP [24].…”