Hybrid Iris Segmentation Method Based on CNN and Principal Curvatures

Abstract

In this article the new hybrid iris image segmentation method based on convolutional neural networks and mathematical methods is proposed. Iris boundaries are found using modified Daugman’s method. Two UNet-based convolutional neural networks are used for iris mask detection. The first one is used to predict the preliminary iris mask including the areas of the pupil, eyelids and some eyelashes. The second neural network is applied to the enlarged image to specify thin ends of eyelashes. Then the principal curvatures method is used to combine the predicted by neural networks masks and to detect eyelashes correctly. The pro- posed segmentation algorithm is tested using images from CASIA IrisV4 Interval database. The results of the proposed method are evaluated by the Intersection over Union, Recall and Precision metrics. The average metrics values are 0.922, 0.957 and 0.962, respectively. The proposed hy- brid iris image segmentation approach demonstrates an improvement in comparison with the methods that use only neural networks.

Fractional Wavelet Transform Phase for Iris Image Key Points Matching

“…6 show that the fractional phase congruence measure can find the image features of the original image with the parameter 𝑎 close to zero. free from the eyelashes, eyelids and glares are found (Tikhonova, 2020). Then the image contrast enhancement is applied.…”

Fractional Fourier Transform Phase for Image Matching

“…To evaluate the order 𝑎𝑎 for calculating the fractional POC-function we use the normalized iris images from the CASIA-IrisV4-Interval database [19]. Figure 8 illustrates the preprocessing of iris images [25].The algorithm detects iris pupil, eyelashes, eyelids, and then the iris image is normalized into a fixed-size rectangle image, then the contrast enhancement is provided. The example of two normalized iris images of one eye is shown in Figure 9.…”

Image Key Point Matching by Phase Congruency

“…For testing, we have used the eye images from the database CASIA-IrisV4-Interval [40]. Figure 5 illustrates the preprocessing of iris images [41]. The algorithm detects the iris edges, the eyelashes, the eyelids, and then normalizes the image (the iris is mapped into a fixed-size rectangle) and performs intensity equalization and contrast enhancement.…”