Note on CASIA-IrisV3 1. Introduction With fast development of iris image acquisition technology, iris recognition is expected to become a fundamental component of modern society, with wide application areas in national ID card, banking, e-commerce, welfare distribution, biometric passport, and forensics, etc. Since 1990s, research on iris image processing and analysis has achieved great progress. However, performance of iris recognition systems in unconstrained environments is still far from perfect. Iris localization, nonlinear normalization, occlusion segmentation, liveness detection, large-scale identification and many other research issues all need further investigation. The success of investigations into such issues often depends on the availability of carefully designed iris image databases of sufficient size. Such publicly available datasets are however very limited. Therefore we are pleased to release to the public domain CASIA Iris Image Database V3.0 (or CASIA-IrisV3 for short) in order to further promote research and progress on iris recognition. 2. Brief Descriptions and Statistics of the Database CASIA-IrisV3 includes three subsets which are labeled as CASIA-IrisV3-Interval, CASIA-IrisV3-Lamp, CASIA-IrisV3-Twins. CASIA-IrisV3 contains a total of 22,035 iris images from more than 700 subjects. All iris images are 8 bit gray-level JPEG files, collected under near infrared illumination. Some statistics and features of each subset are summarized in Table 1. Almost all subjects are Chinese except a few in CASIA-IrisV3-Interval. Because the three data sets were collected in different times, only CASIA-IrisV3-Interval and CASIA-IrisV3-Lamp have a small overlap in subjects. 2.1 CASIA-IrisV3-Interval and CASIA V1.0 Iris images of CASIA-IrisV3-Interval were captured with our self-developed iris camera (Fig.1a). CASIA-IrisV3-Interval is a superset of CASIA V1.0 which has been requested by and released to more than 1,500 researchers/teams from 70 countries and regions (as of June 2006). CASIA V1.0 contains 756 iris images from 108 subjects. In order to protect our IPR in the design of our iris camera (especially the NIR illumination scheme), the pupil regions of all iris images in CASIA V1.0 were automatically detected and replaced with a circular region of constant intensity to mask out the specular reflections from the NIR illuminators (see Fig. 2). Such editing clearly makes iris boundary detection much easier but has minimal or no effects on other components of
an iris recognition system, such as feature extraction and classifier design. As patents have been granted to us on the design of the iris camera, we are now happy to release the original unmasked images. The availability of CASIA-IrisV3-Interval may make CASIA V1.0 obsolete. 2.2 CASIA-IrisV3-Lamp and CASIA-IrisV3-Twins Both CASIA-IrisV3-Lamp and CASIA-IrisV3-Twins were collected using OKI s hand-held iris sensor (Fig.1b). A lamp was turned on/off close to the subject to introduce more intra-class variations when we collected CASIA-IrisV3-Lamp. CASIA-IrisV3-Twins contains iris images from 100 pairs of twins. Table 1 Statistics of CASIA-IrisV3 Characteristics CASIA-IrisV3-Interval CASIA-IrisV3-Lamp CASIA-IrisV3-Twins Sensor Self-developed OKI s IRISPASS-h OKI s IRISPASS-h Environment Indoor Indoor with lamp on/off Outdoor Session Most of the images were captured in two sessions, with at one one least one month interval No. of subjects 249 411 200 No. of classes 395 819 400 No. of images 2639 16213 3183 Resolution 320*280 640*480 640*480 Features Very good image Nonlinear The first publicly quality with extremely deformation due to available twins iris clear iris texture variations of visible image dataset details illumination Total A total of 22,035 iris images from more than 700 subjects and 1500 eyes
(a) Iris camera developed at CASIA (b) Iris camera from OKI Figure 1 Iris image sensors used in CASIA V3.0 construction
(a) An image from CASIA V1.0 (b) An image from CASIA-IrisV3-Interval Figure 2 Example images from CASIA V1.0 and CASIA-IrisV3-Interval. (a) A sample iris image from CASIA V1.0; (b) A sample iris image from CASIA-IrisV3-Interval. 3. Image Formats and Download Instructions The images of the first two subsets are stored as: SubsetName\YYY\E\SXYYYENN.jpg X: the index of subset YYY: the unique identifier of subject in each subset. E: L denotes left eye and R denotes right eye NN: the index of image in the class.
The images of CASIA-IrisV3-Twins are stored as: CASIA-IrisV3-Twins\XX\YE\S3XXYENN.jpg XX: the index of family Y: the identifier to one of twins E: L denotes left eye and R denotes right eye NN: the index of image in the class. Researchers requesting this database should follow the following steps: Download the application form at the website: http://www.cbsr.ia.ac.cn/irisdatabase.htm. Fill the application form. Send the form via email to casia_iris@nlpr.ia.ac.cn. Check your email to find a login account and a password of our website after one day, if your application has been approved. Download the CASIA-IrisV3 from our website with the authorized account within 48 hours. 4. Copyright Note and Contacts The database is released for research and educational purposes. We hold no liability for any undesirable consequences of using the database. All rights of the CASIA database are reserved. Any person or organization is not permitted to distribute, publish, copy, or disseminate this database. In all documents and papers that report experimental results based on this database, our efforts in constructing the database should be acknowledged as: Portions of the research in this paper use the CASIA-IrisV3 collected by the Chinese Academy of Sciences Institute of Automation (CASIA) and a reference to "CASIA-IrisV3, http://www.cbsr.ia.ac.cn/irisdatabase.htm" should be included. A copy of all reports and papers that are for public or general release that use the CASIA-IrisV3 should be forwarded upon release or publication to Professor Tieniu Tan Center for Biometrics and Security Research National Laboratory of Pattern Recognition Institute of Automation, Chinese Academy of Sciences P.O.Box 2728 Beijing 100190 China or send electronic copies to znsun@nlpr.ia.ac.cn. Questions regarding this database can be addressed to Dr. Zhenan Sun at
Dr. Zhenan Sun Center for Biometrics and Security Research National Laboratory of Pattern Recognition Institute of Automation, Chinese Academy of Sciences P.O.Box 2728 Beijing 100190 China Tel: +86 10 8261 0278 Fax: +86 10 6255 1993 Email: znsun@nlpr.ia.ac.cn