Iris Recognition: An Entropy-Based Coding Strategy Robust to Noisy Imaging Environments
The iris is currently accepted as one of the most accurate traits for biometric purposes. However, for the sake of accuracy, iris recognition systems rely on good quality images and significantly deteriorate their results when images contain large noisy r
- PDF / 1,531,551 Bytes
- 12 Pages / 430 x 660 pts Page_size
- 90 Downloads / 231 Views
ct. The iris is currently accepted as one of the most accurate traits for biometric purposes. However, for the sake of accuracy, iris recognition systems rely on good quality images and significantly deteriorate their results when images contain large noisy regions, either due to iris obstructions (eyelids or eyelashes) or reflections (specular or lighting). In this paper we propose an entropy-based iris coding strategy that constructs an unidimensional signal from overlapped angular patches of normalized iris images. Further, in the comparison between biometric signatures we exclusively take into account signatures’ segments of varying dimension. The hope is to avoid the comparison between components corrupted by noise and achieve accurate recognition, even on highly noisy images. Our experiments were performed in three widely used iris image databases (third version of CASIA, ICE and UBIRIS) and led us to observe that our proposal significantly decreases the error rates in the recognition of noisy iris images.
1 Introduction Continuous efforts have been made in searching for robust and effective iris coding methods, since Daugman’s pioneering work on iris recognition was published. Iris recognition has been successfully applied in such distinct domains as airport checkin or refugee control. However, for the sake of accuracy, current systems require that subjects stand close (less than two meters) to the imaging camera and look for a period of about three seconds until the data is captured. This cooperative behavior is indispensable to capture images with enough quality to the recognition task. Simultaneously, it restricts the range of domains where iris recognition can be applied, namely within heterogeneous lighting conditions or under natural lighting environments. In this context, the overcome of these imaging constrains has motivated the efforts of several authors and deserves growing attention from the research community. Although some of the published iris recognition algorithms perform a noise detection stage and produce a binary mask - used to avoid that noisy components of the biometric signatures are taken into account - we believe that highly heterogeneous lighting environments (specially under natural light) lead to the appearance of regions which, even for humans, are very difficult to classify as ”noisy” or ”noise-free”. Figure 1 illustrates some of the noise factors that result of less constrained image capturing environments. In figure 1b large iris regions obstructed by reflections (lighting and specular) can be observed, some of them very difficult to distinguish from the noise-free ones. G. Bebis et al. (Eds.): ISVC 2007, Part I, LNCS 4841, pp. 621–632, 2007. c Springer-Verlag Berlin Heidelberg 2007
622
H. Proenc¸a and L.A. Alexandre
(a) Iris image with good quality.
(b) Noisy iris image.
Fig. 1. Comparison between a good quality image and a noise-corrupted one. Figure 1a was captured under high constrained imaging conditions and is completely noise-free. Oppositely, figure 1b incorporat
Data Loading...