Texture code matrix-based multi-instance iris recognition

Abstract

This paper proposes a novel texture feature for iris recognition. The iris recognition system consists of three major components: pre-processing, feature extraction and classification. During pre-processing, iris is segmented using constrained circular Hough transform, which reduces both time and space complexity. In this work, from normalized iris image, a novel texture code matrix is generated, which is then used to obtain a co-occurrence matrix. Finally, desired texture features are computed from this cooccurrence matrix. Here, a two-class classification technique is adopted to develop a multi-class multimodal biometric system using fusion. The performance of the proposed system is tested on four standard iris image databases, namely UPOL, CASIA-Iris V3 Interval, MMU1 and IITD, which shows the efficacy of the proposed feature.

A multi-biometric iris recognition system based on a deep learning approach

“…In term of the Rank-1 identification rate, the highest results were obtained by the proposed system using these two databases. Although Umer et al [57] also achieved a 100% recognition rate for the CASIA-Iris-V3 database, the proposed system achieved a better running time to establish the person's identity from 120 persons from the same database instead of 99 persons as in [57]. In addition, they obtained inferior results for the IITD database in terms of both recognition rate and running time.…”

“…After that, larger numbers of epochs were also investigated using the same training methodology, including 200, 300, 400, 500 and 600 epochs. The CMC curves shown [54] 99.18 ---Uhl et al [55] 74.00 0.21 --Ugbaga et al [56] 98.90 ---Umer et al [57] 95.87 0.89 98.48 0.77 Wild et al [58] 98.13 -97.60 -Aydi et al [59] 96.51 9.049 --Pawar et al [60] 96.88 ---Mehrotra et al [61] 99. obtained on the validation set. Therefore, 500 epochs were taken as the initial number of epochs in our assessment procedure for all remaining experiments, since the learning process still achieved good generalization without overfitting.…”

“…In this work, the running time was measured by implementing the proposed approaches using a laboratory in Bradford University consisting of 25 PCs with the Windows 8.1 operating system, Intel Xeon E5-1620 CPUs and 16 GB of RAM. The system code was written to run in MATLAB [67] 98.45 -Zhang and Guan [68] 99.60 -Roy et al [69] 97.21 0.995 Li et al [70] 99.91 -Bharath et al [71] 84.17 0.44 IrisConvNet System 100 0.89 IITD Umer et al [57] 99.52 1.11 Bharath et al [71] 95.93 0.10 Nalla and Chalavadi [72] 86.00 -Elgamal and Al-Biqami [73] 99.5 -Minaee et al [74] 99.20 -Dhage et al [75] 97.81 93.24 Abhiram et al [76] 97.12 -IrisConvNet system 100 0.81…”

“…Radon transform + gradient-based isolation Euclidean distance 4:1 (training:testing) De Costa and Gonzaga [66] Dynamic features Euclidean distance Cross-validation Dhage et al [75] DWT + DCT Euclidean distance 9:1 (training:testing) Elgamal and Al-Biqami [73] DWT + PCA KNN -Kerim et al [65] Co-occurrence matrix Euclidean distance -Li et al [70] ALBP KNN + SVM 4:1 (training:testing) Ma et al [63] Circular symmetric filter Nearest feature line 3:2 (training:testing) Minaee et al [74] Scattering transform Minimum distance Cross-validation Ng et al [67] Haar wavelet transform Hamming distance -Nalla and Chalavadi [72] Log-Gabor wavelet Online Dictionary Learning Cross-validation Roy et al [69] Multi-perturbation Shapley analysis SVM Cross-validation Umer et al [57] TCM with ordered PB SVM + Fusion Leave-one-out Vatsa et al [64] Gabor transform and euler numbers Mahalanobis distance Cross-validation Zhang and Guan [68] Empirical mode decomposition KNN -IrisConvNet system Convolutional Neural Network Softmax classifier + fusion Cross-validation…”

Cork as a Unique Object: Device, Method, and Evaluation

“…Texture code matrix [50] 99.96 0.00 --Triplet Half-Band Filter Bank (THFB) + k-out-of-n [51] 99 Table 4. Results from the MMU1 database [3].…”