A Low-quality Iris Image Segmentation Algorithm Based on SRN-UNet

Huijuan TIAN; Jiahao ZHAI; Jianxin LIU; Jiawei LIU; Linlin DENG

doi:10.3788/gzxb20225102.0210006

Acta Photonica Sinica, Volume. 51, Issue 2, 0210006(2022)

A Low-quality Iris Image Segmentation Algorithm Based on SRN-UNet

Huijuan TIAN^1,2、*, Jiahao ZHAI^1,2, Jianxin LIU³, Jiawei LIU^1,2, and Linlin DENG^1,2

¹Tianjin Key Laboratory of Optoelectronic Detection Technology and System，School of Electrical and Electronics Engineering，Tiangong University，Tianjin 300387，China

²Engineering Research Center of Ministry of Education on High Power Solid State Lighting Application System，Tianjin 300387，China

³Tianjin Chengke Transmission Electromechanical Technology Co．，Ltd．，Tianjin 300384，China

show less

Abstract Get PDF(in Chinese)

Figures & Tables(12)

Fig. 1. Overall network structure of SRN-UNet

Download full size

View in Article

Fig. 2. Network structure of SE-ResNext module

Download full size

View in Article

Fig. 3. The basic unit of ResNet and ResNext

Download full size

View in Article

Fig. 4. Network structure of SENet

Download full size

View in Article

Fig. 5. Data set sample display

Download full size

View in Article

Fig. 6. Part of the data set and corresponding annotations

Download full size

View in Article

Fig. 7. The prediction results of different algorithms for near-infrared iris images

Download full size

View in Article

Fig. 8. The prediction results of different algorithms for low-quality iris images

Download full size

View in Article

Table 1. Comparison of segmentation performance of different loss functions
View table
View in Article
Table 1. Comparison of segmentation performance of different loss functions
Loss Miou/% F1 score/% Precision/%
Cross entropy loss 93.88 96.84 97.08
Dice+Focal 95.19 97.48 97.82

Table 2. Time required for network operation（s）
View table
View in Article
Table 2. Time required for network operation（s）
Algorithm Batch=10 Batch=20 Batch=50
U-Net 0.909 4 1.749 3 4.414 7
Ours 0.621 0 1.237 4 3.048 3

Table 3. Comparison of different iris segmentation algorithms
View table
View in Article
Table 3. Comparison of different iris segmentation algorithms
Algorithm Miou/% F1 score/% Precision/%
U-Net 90.99 95.21 92.44
SegNet 91.18 95.32 92.15
PSPNet 92.28 95.97 94.12
Deeplabv3 92.31 96.00 95.73
Ours 95.19 97.48 97.82

Table 4. Influence of each module on whole model
View table
View in Article
Table 4. Influence of each module on whole model
U-Net ResNext-UNet SENet-UNet Miou/% F1 score/% Precision/%
＋ 91.02 95.20 92.51
＋＋ 91.89 95.47 93.18
＋＋ 92.01 95.30 93.27
＋＋＋ 95.19 97.48 97.82

Tools

Get Citation

Copy Citation Text

Huijuan TIAN, Jiahao ZHAI, Jianxin LIU, Jiawei LIU, Linlin DENG. A Low-quality Iris Image Segmentation Algorithm Based on SRN-UNet[J]. Acta Photonica Sinica, 2022, 51(2): 0210006

Download Citation

EndNote(RIS)BibTex Plain Text

Set citation alerts for article

Save article for my favorites