Adaboost and multi-orientation 2D Gabor-based noisy iris recognition

doi:10.1016/j.patrec.2011.08.014

Pattern Recognition Letters

Volume 33, Issue 8, 1 June 2012, Pages 978-983

https://doi.org/10.1016/j.patrec.2011.08.014 Get rights and content

Abstract

In this paper, we present a noisy iris recognition frame which is learned by Adaboost on a 2D Gabor-based feature set. First, the irises are segmented and normalized by rubber sheet or simplified rubber sheet according to whether segmentations are accurate or not. Then, irises are divided into different amount of patches according to normalization. Moreover, a feature set is constructed based on 2D-Gabor for whole iris and patches. Finally, Adaboost learning is used for accurately and inaccurately segmented irises separately.

The proposed method was evaluated by the NICE:II (Noisy Iris Challenge Evaluation – Part 2). We were ranked 2nd among all of the 67 participants from 29 different countries/districts.

Highlights

► Both global and local texture information are used for iris recognition. ► A set of regional hamming distance combined with Adaboost is used for classifiers. ► A new method is proposed for the not accurately segmented noise irise recognition. ► Two recognition methods are adopted according to the segmentations.

Introduction

Iris recognition is one of the most important biometrics. Since Daugman proposed the first iris recognition system based on 2D Gabor (Daugman, 1993), researchers have devoted great efforts to iris recognition (Daugman, 2007, Li et al., 2004, Ma et al., 2003, Proenca and Alexandre, 2007b, Sun and Tan, 2009, Wildes, 1997).

Noisy iris recognition is quite challenging. These images are captured in a natural environment and in visible light, without cooperation and illumination controlled, even at a distance or on-the-move.

This paper presents a noisy iris recognition method, which has been ranked 2nd in NICE:II contest. It is organized as follows: In Section 2, the proposed method of noisy iris recognition is presented. In Section 3, technical details and results of experiments are described. Finally, some conclusions are drawn in Section 4.

Section snippets

Proposed method

The proposed noisy iris recognition method mainly includes four steps: First, irises are segmented and the segmentations are evaluated. Second, they are normalized by rubber sheet or simplified rubber sheet according to whether segmentations are accurate or not. Then, a feature set is constructed based on 2D-Gabor. At last, Adaboost learning is used for accurately and inaccurately segmented irises separately for recognition.

Experiments and performance

The training database is composed of 1000 noisy iris images in NICE:II. We select 810 images from the training database to carry out experiments. These 810 iris images are used to train both Adaboost_AS and Adaboost_IAS. The details and results of experiments are described in this section.

Conclusion

In this paper, a noisy iris recognition method is proposed. The method provides a framework which combines regional Gabor feature with Adaboost algorithm for noisy iris recognition. With this framework, two different recognition strategies are designed according to the accuracy of pupil segmentations.

The proposed method has been validated in NICE:II contest. It was ranked 2nd among all of the 67 participants from 29 different countries/districts.

Acknowledgements

The authors would like to thank SOCIA Lab (Soft Computing and Image Analysis Group) of University of Beira Interior in Portugal for their great dedication in organizing NICE:II. We thank Dr. Yang Guo for his excellent advice and code of RANSAC. We are grateful to Hegui Zhu and Lianping Yang for their great help in language. We also appreciate the referees for their constructive recommendations very much.

References (20)

J. Daugman
The importance of being random: Statistical principles of iris recognition
Pattern Recognition
(2003)
P.H. Li et al.
Robust and accurate iris segmentation in very noisy iris images
Image Vision Comput.
(2010)
X.C. Yin et al.
Feature combination using boosting
Pattern Recognition Lett.
(2005)
J. Daugman
Statistical richness of visual phase information: Update on recognizing persons by iris patterns
Internat. J. Comput. Vision
(2001)
J. Daugman
How iris recognition works
IEEE Trans. Circuits Systems Video Technol.
(2004)
J. Daugman
New methods in iris recognition
IEEE Trans. Systems Man Cybernet. Part B: Cybernet.
(2007)
J.G. Daugman
High confidence visual recognition of persons by a test of statistical independence
IEEE Trans. Pattern Anal. Machine Intell.
(1993)
Z.F. He et al.
Efficient iris spoof detection via boosted local binary patterns
Z.F. He et al.
Toward accurate and fast iris segmentation for iris biometrics
IEEE Trans. Pattern Anal. Machine Intell.
(2009)
M. Li et al.
Efficient iris recognition by characterizing key local variations
IEEE Trans. Image Process.
(2004)

There are more references available in the full text version of this article.

Cited by (66)

Effectiveness evaluation of infrared camouflage using image saliency
2018, Infrared Physics and Technology
This paper reports on an alternative approach to the effectiveness evaluation of IR camouflage using image saliency. The master saliency and interior saliency of a target were computed according to its background-related features and internal shape features, respectively. The fusion of master saliency and interior saliency were further used to obtain a comprehensive saliency value, by which the camouflage effectiveness value of a target was finally obtained. Under the same experimental conditions, the target with camouflage matching well with the background had the highest camouflage effectiveness value, while the target without camouflage had the lowest camouflage effectiveness value, indicating that such a value can well reflect the camouflage effectiveness of different targets in a certain background. Acting as a quantitative index, the camouflage effectiveness value makes it possible for us to directly differentiate the camouflage effectiveness of not only different targets under the same conditions but also one target under different conditions, which should be useful for the development of IR camouflage materials and the deployment of military equipment in battlefield.
Deep periocular representation aiming video surveillance
2018, Pattern Recognition Letters
Citation Excerpt :
All the four approaches are combined with sum-rule at matching score level. Wang et al. [39] proposed a method using only iris modality. First, the iris is normalized as proposed in [40] and partitioned into several segments.
Usually, in the deep learning community, it is claimed that generalized representations that yielding outstanding performance / effectiveness require a huge amount of data for learning, which directly affect biometric applications. However, recent works combining transfer learning from other domains have surmounted such data application constraints designing interesting and promising deep learning approaches in diverse scenarios where data is not so abundant. In this direction, a biometric system for the periocular region based on deep learning approach is designed and applied on two non-cooperative ocular databases. Impressive representation discrimination is achieved with transfer learning from the facial domain (a deep convolutional network, called VGG) and fine tuning in the specific periocular region domain. With this design, our proposal surmounts previous state-of-the-art results on NICE (mean decidability of 3.47 against 2.57) and MobBio (equal error rate of 5.42% against 8.73%) competition databases.
Certain investigation on iris image recognition using hybrid approach of Fourier transform and Bernstein polynomials
2017, Pattern Recognition Letters
Citation Excerpt :
Tieniu et al. [19] proposed a robust iris recognition using novel region growing method. Qi Wang et al. [20] proposed an iris recognition using 2D Gabor filter and Adaboost learning. Gabor filter is used to extract the iris features and Adaboost learning is used for classification.
In the era of biometric research the recent developments in real time biometric world, iris recognition is considered as one of the most important approach used in person identification based biometric authentication. This approach is considered than the other biometric authentication methods such as behavioral biometrics which specially includes Keystroke, Speaker Recognition whereas the another side of biometric authentication is Physical biometrics which specially includes Fingerprint Recognition, Voice Recognition, Finger Geometry Recognition and Facial Recognition. From these considerations, the hybrid approach of Fourier transform and Bernstein polynomial for iris recognition has been proposed. The novelty of this proposed method results in improving the Accuracy, False Acceptance Rate (FAR), False Rejection Rate (FRR) and Equal Error Rate (EER) for the given iris image. In addition to this, Singular Value Decomposition (SVD) is used for iris image pre-processing. Circular Hough transform (CHT) and Canny edge detection (CED) are applied for iris image segmentation which segments the individual region of the input images. After the image is segmented, Fourier transform and Bernstein polynomial have been applied to extract the features from the segmented iris image, which is the most important step for obtaining the texture details, which are independent and uncorrelated even for identical pairs. Support Vector Machine (SVM) is used for image classification. Perhaps, Feature extraction and Classification of iris image are mainly based on the rich texture details present in iris image. Finally, our proposed system is applied on UBIRIS database and our research experiment provides better accuracy and recognition rates compared to the combined iris recognition techniques such as Fourier transform with SVM, Bernstein polynomial with SVM, Fourier transform with KSVM, Bernstein polynomial with KSVM and hybridization of Fourier and Bernstein polynomial with SVM.
Second-order convolutional networks for iris recognition
2022, Applied Intelligence
Ocular recognition databases and competitions: a survey
2022, Artificial Intelligence Review
On measuring and employing texture directionality for image classification
2021, Pattern Analysis and Applications

View all citing articles on Scopus

View full text

Adaboost and multi-orientation 2D Gabor-based noisy iris recognition

Abstract

Highlights

Introduction

Section snippets

Proposed method

Experiments and performance

Conclusion

Acknowledgements

Pattern Recognition

Image Vision Comput.

Pattern Recognition Lett.

Statistical richness of visual phase information: Update on recognizing persons by iris patterns

Internat. J. Comput. Vision

How iris recognition works

IEEE Trans. Circuits Systems Video Technol.

New methods in iris recognition

IEEE Trans. Systems Man Cybernet. Part B: Cybernet.

High confidence visual recognition of persons by a test of statistical independence

IEEE Trans. Pattern Anal. Machine Intell.

Efficient iris spoof detection via boosted local binary patterns

Toward accurate and fast iris segmentation for iris biometrics

IEEE Trans. Pattern Anal. Machine Intell.

Efficient iris recognition by characterizing key local variations

IEEE Trans. Image Process.