An experimental study of relative total variation and probabilistic collaborative representation for iris recognition
- PDF / 1,632,137 Bytes
- 19 Pages / 439.37 x 666.142 pts Page_size
- 8 Downloads / 212 Views
An experimental study of relative total variation and probabilistic collaborative representation for iris recognition Pradeep Karn 1
1
1
& XiaoHai He & Jin Zhang & Yanteng Zhang
1
Received: 21 May 2019 / Revised: 29 May 2020 / Accepted: 21 July 2020 # Springer Science+Business Media, LLC, part of Springer Nature 2020
Abstract
Iris images collected under different conditions often suffer from specular reflections, cast shadows, motion blur, defocus blur, occlusion caused by eyelashes and eyelids, eyeglasses, hair and other artifacts. Existing iris recognition systems do not perform well on these types of images. To overcome these problems, an iris recognition method based on relative total variation (RTV) and probabilistic collaborative representation is proposed. RTV uses the l1 norm regularization method to robustly suppress noisy pixels to achieve accurate iris localization, while probability collaborative representation maximizes the probability that the test sample belongs to each of the multiple classes. The final recognition rate is calculated based on the class having maximum probability. Experimental results using CASIA-V4Lamp and IIT-Delhi V1iris image databases showed that the proposed method achieved competitive performance in both recognition accuracy and computational efficiency. Keywords Biometrics . Iris recognition . Relative total variation . Gabor filter . Probabilistic collaborative representation-based classification
1 Introduction Biometrics [21] is an individual’s trait that can be used for personal identification. Biometrics is used in a variety of applications. The most common biometrics recognition systems include face, fingerprints, voice patterns, Deoxyribonucleic acid (DNA), hand vein geometry, signature and iris patterns [43]. Among all biometrics systems, iris recognition systems are considered the most reliable and accurate [7] because of their statistically unique features, which remain the same throughout an individual’s life. Although a lot of progress has been
* Pradeep Karn [email protected]
1
Image Information Institute, College of Electronics and Information Engineering, Sichuan University, Chengdu 610065, People’s Republic of China
Multimedia Tools and Applications
made in the development of highly improved iris recognition, iris recognition still faces some challenges in practical scenario, especially if the image acquisition is not constrained [3, 37]. The eye image acquired in unconstrained environments contain noise and complex details, such as eyelid and eyelash occlusion, blur, specular reflection, cast shadows, low illumination and off-axis gaze. These factors will increase intra-class differences, seriously affect the recognition accuracy, and bring many challenges to the feature extraction and matching methods in iris recognition system. To solve these problems, we propose an improved iris recognition method based on relative total variation (RTV) [54, 56] and probabilistic collaborative representation-based classification (ProCRC). This method can accur
Data Loading...
