Fitting an ellipse to the iris boundaries accounts for the projective distortions present in off-axis images of the eye and provides the contour fitting necessary for the dimensionless mapping used in leading iris recognition algorithms. Previous iris segmentation efforts have either focused on fitting circles to pupillary and limbic boundaries or assigning labels to image pixels. This paper approaches the iris segmentation problem by adapting the Starburst algorithm to locate pupillary and limbic feature pixels used to fit a pair of ellipses. The approach is evaluated by comparing the fits to ground truth. Two metrics are used in the evaluation, the first based on the algebraic distance between ellipses, the second based on ellipse chamfer images. Results are compared to segmentations produced by NDIRIS over randomly selected images from the Iris Challenge Evaluation database. Statistical evidence shows significant improvement of Starburst's elliptical fits over the circular fits on which NDIRIS relies. I. INTRODUCTIONExcept for several relatively unique approaches, e.g., [3], [16], common iris segmentation methods model the iris as a pair of circles [5]. Although the inner and outer boundaries of the iris may be roughly approximated by circles, they rarely appear as true circles in images [9]. The iris image is subject to perspective projection. It is approximately planar. Any circle that lies in a plane not fronto-parallel to the camera will appear elliptical in the image plane. The segmentation model must account for such distortions. A general ellipse model is therefore more appropriate than a restricted circular model to compensate for this type of distortion.The Starburst algorithm was introduced by Li, Babcock, and Parkhurst for the purpose of eye tracking [14]. For such an application, Starburst's main objective is to identify feature points on the limbus for subsequent localization of the pupil center. Starburst then fits an ellipse to the limbic pixels, operating under the implicit assumption that the center of that ellipse coincides with the pupil center. The pupil center is then used for estimating the point of gaze, or POG, of a viewer wearing the eye tracking apparatus.In this paper we adapt the Starburst algorithm for the purpose of iris segmentation. The novelty behind our adaptation is the simultaneous identification of both pupillary and limbic boundaries, fitting ellipses to both contours, thereby producing an iris segmentation suitable for subsequent iris recognition as well as eye tracking applications. Such contour fitting is an essential component of iris recognition [8].

Figure 1: Ellipse fitting: detection of features on both pupil and iris boundaries; ellipses fit to sets of 5 randomly selected points; luminancebased delineation of features into two sets; proper detection of erroneous ellipse spanning both pupil and iris. AbstractWe present a low-cost wearable eye tracker built from off-the-shelf components. Based on the open source openEyes project (the only other similar effort that we are aware of), our eye tracker operates in the visible spectrum and variable lighting conditions. The novelty of our approach rests in automatically switching between tracking the pupil/iris boundary in bright light to tracking the iris/sclera boundary (limbus) in dim light. Additional improvements include a semi-automatic procedure for calibrating the eye and scene cameras, as well as an automatic procedure for initializing the location of the pupil in the first image frame. The system is accurate to two degrees visual angle in both indoor and outdoor environments.

No abstract