1.       Introduction

Near-Infra-Red (NIR) light (in the range of 780 nm to 840 nm) is capable of effectively capturing the iris pattern since light in this range is scattered in the internal structures of the iris regardless of the color it is, or the possible low contrast between the iris and the pupil in those individuals with dark irises. However, most commercial sensors, such as video surveillance cameras, do not have NIR sensors to perform this type of capture. On the other hand, the rise of mobile devices such as smart phones and their integrated cameras are already used for various biometric applications. Nevertheless, in the case of iris biometry this can be hampered by the limiting factor of not having NIR sensors. Therefore, if you intend to use a sensor that works in the visible spectrum (in the range of 380 nm to 720 nm) to capture iris patterns, the success could be limited only to those instances of light color iris and that are captured in a controlled scenario. In view of the growing popularity of iris biometry based on this type of sensor [1], it is important to address this problem due to the wide spectrum of applications that can be developed. The acquisition of video-based eye images for iris recognition is an interesting alternative in the current context of the extensive use of mobile devices and video surveillance cameras [2,3]. This modality can provide more information from video capture of eye region.

The problem in these systems is the generated large amount of information from the video capture and how to decide what information will be passed to the system in order to perform the recognition process. A metric for evaluating the quality of eye images combined with automatic image detection can be an alternative. In this work, an approach for video iris recognition is proposed; it is based on a scheme whose novelty is in the possibility of evaluating the quality of the eye image in real time simultaneously with process of video capture. For this purpose, a measure of eye image quality is proposed, it takes into account the elements defined in the ISO/ IEC 19794-6: 2005 standard [4]. The combination of the proposed measure with automatic eye detection method ensures that eye images are extracted so that they do not have elements that negatively influence the identification process such as closed eyes and out-of-angle look. The work is structured as fallows. Section 2 discuss the related works, section 3 presents the proposed approach, in section 4 the experimental results are presented and discussed, and finally the conclusions of the work are set.

2.       Related Works

Evaluating the quality of iris images is one of the recently identified topics in the field of iris biometry [5,6]. In general, quality metrics are used to decide whether the image should be discarded or processed by the iris recognition system. The quality of iris images is determined by many factors depending on the environmental and camera conditions and on the person, being identified [5]. Some of the quality measures reported in literature [6] focus on the evaluation of iris images after the segmentation process, which makes the systems in their capture stage, allow the processing of poor and good quality images. The main lack of these approaches is that the evaluation of the iris image quality is reduced to the estimation of a single or a couple of factors [3], such as out-of-focus blur, motion blur, and occlusion. Other authors [6,7] use more than three factors to evaluate the quality of the iris image: such as the degree of defocusing, blurring, occlusion, specular reflection, lighting, out of angle. Its main lack is they consider that the degradation of some of the estimated parameters below the threshold brings to zero (veto power) the measure that integrates all the evaluated criteria. This may be counterproductive in some systems where the capture conditions are not optimal.

The ISO / IEC 19794-6: 2005 [4] standard identified several properties of the iris image that influence the recognition accuracy. These factors include the distance of the acquisition system from the user, the pixel density of the iris texture and the degree of image blurring. In practice, some of these factors can be controlled by the correct selection of the camera, the correct analysis of the Depth of Field (DOF) and the Field of Vision (FOV). A quality measure that considers the parameters established in the standard [4] and evaluates detected eye image before the segmentation can produce a reduction in errors in the next steps of the system with a consequent increase in recognition rates.

3.       The Proposed Approach

Figure 1 shows the general scheme of the proposed approach. The novelty of the proposed approach lies in the proposal of a new quality metric and it combination with a previous stage of eye image detection. This approach will ensure that the detected eye images do not have elements that negatively influence the identification related with: illumination, sharpness, blurring, gaze, occlusion, pixel density of image.

3.1.  Iris Video Capture and Eye Detection

In [9] the authors perform an analysis of the implication of using iris images in the Visible Spectrum (VS). They demonstrated how the use of a white LED light source positively influences the recognition rates of an iris recognition system. In our proposal, these precepts using a similar design to capture the video were assumed in this work. Detection of eye images is achieved through the classical Viola and Jones algorithm [10]. A detector was trained to detect open eyes containing pupils and iris with or without specular light reflection. The training set consisted of 1000 labeled eye images taken from the MobBio [11], UTIRIS [12] databases and our own dataset (see section 4). As positive samples were taken images of open eyes looking forward and as negative samples images of blurred, closed or occluded eyes.

3.2.  Image Quality Evaluation

Among the parameters established by the standard [4], the FOV and the focal length are two parameters required to determine the distance between the subject and the camera. The FOV indicates the optimal distance between the subject and the camera for a given pixel density and the focal length is the zoom of the subject in the image. The FOV can be calculated by equation 1.

Where (

Figure 1: General scheme of the proposed approach.

Figure 2: Samples of eye images with Qindex>=1(above) and Qindex<1(below).

Figure 3: DET curves obtained in the experiment for the own video dataset capturing 1000 images containing all classes.

1.       Raja KB, Raghavendra R, Vemuri VK, Busch C (2015) Smartphone based visible iris recognition using deep sparse filtering. Pattern Recognition Letters 57: 33-42.

2.       Hollingsworth K, Peters T, Bowyer K (2009) Iris recognition using signal-level fusion of frames from video, IEEE Trans. Inf. Forensics Secur 4: 837-848.

3.       Garea-Llano E, García-Vázquez M, Colores -Vargas JM, Zamudio-Fuentes LM, Ramírez-Acosta AA (2018) Optimized robust multi-sensor scheme for simultaneous video and image iris recognition. Pattern Recognition Letters 101: 44-45.

4.       ISO/IEC        19794-6:2005      Part         6:            Iris          image     data",     ISO.

5.       Schmid N, Zuo J, Nicolo F, Wechsler H (2016) Iris Quality Metrics for Adaptive Authentication. In: Bowyer KW, Burge MJ. Handbook of Iris Recognition (2^nd Edition). Advances in Computer Vision and Pattern Recognition, Springer-Verlag London (101-118).

6.       Daugman J, Downing C (2016) Iris Image Quality Metrics with Veto Power and Nonlinear Importance Tailoring. In: Rathgeb C, Busch C. Iris and Periocular Biometric Recognition. IET Pub. Pg No: 83-100.

7.       Zuo J, Schmid NA (2008) An automatic algorithm for evaluating the precision of iris segmentation. In: International Conference on Biometrics: Theory, Applications, and Systems (BTAS’08), Washington, DC, USA.

8.       Bowyer KW, Hollingsworth K, Flynn PJ (2008) Image understanding for iris biometrics: a survey. J Comput Vis Image Underst 110: 281-307.

9.       Raja KB, Raghavendra R, Busch C (2015) Iris Imaging in Visible Spectrum using White LED. 7th International Conference on Biometrics Theory, Applications and Systems (BTAS), IEEE.

10.    Viola P, Jones M (2004) Rapid Object Detection Using a Boosted Cascade of Simple Features. Mitsubishi Electric Research Laboratories, Inc., 2004 201 Broadway, Cambridge, Massachusetts 02139.

11.    Monteiro C, Oliveira HP, Rebelo A, Sequeira AF (2013) Mobbio 2013: 1st biometric recognition with portable devices competition.

12.    Hosseini MS, Araabi BN, Soltanian-Zadeh H (2010) Pigment Melanin: Pattern for Iris Recognition. IEEE Transactions on Instrumentation and Measurement 59: 792-804.

13.    Rathgeb C, Uhl A, Wild P (2012) Iris Recognition: From Segmentation to Template Security. In: Advances in Information Security 59. Springer, New York.

14.    Lowe DG (2004) Distinctive image features from scale-invariant key points. Int J Comput Vis 60: 91-110.

15.    SOCIA Lab (2012) University of Beira Interior: UBIRIS.v1 Database.