EAI-NET: Effective and Accurate Iris Segmentation Network

• Published in: Pattern Recognition and Machine Intelligence pp. 442 - 451
• Main Authors: Rajpal, Sanyam, Sadhya, Debanjan, De, Kanjar, Roy, Partha Pratim, Raman, Balasubramanian
• Format: Book Chapter
• Language: English
• Published: Cham Springer International Publishing
• Series: Lecture Notes in Computer Science
• Subjects: Iris; Segmentation; U-Net
• ISBN: 3030348687; 9783030348687
• ISSN: 0302-9743; 1611-3349
• DOI: 10.1007/978-3-030-34869-4_48

In iris-based biometric models, segmentation of the iris region from the rest of the eye is a crucial step. The quality of the segmented region directly affects the extracted iris features, which subsequently determines the overall recognition accuracy of the model. In this work, we propose EAI-Net, which is an effective and accurate iris segmentation network based on the U-Net architecture. In comparison to the previous works, we treat the segmentation process as a 3-class problem wherein the pupil, iris and the rest of the image are treated as separate classes. Furthermore, we have increased the complexity degree of our model by encoding the complex regions of the iris more efficiently. We have conducted both qualitative and quantitative assessments of our results over four benchmark iris databases - UBIRISv2, IITD, CASIAv4-Interval, and CASIAv4-Thousand. The obtained results demonstrate the superiority of our model over the other state-of-the-art deep-learning based approaches in solving the problem of iris segmentation in both the visible (VIS) and near-infrared (NIR) spectrum.