International Conference on Computer and Knowledge Engineering

Home / 11th International Conference on Computer and Knowledge Engineering

Iris Detection and Segmentation Using Deep Learning

Authors :

Ali Khaki¹ Ali Aghagolzadeh² Bagher Rahimpour Cami³

1- Faculty of Computer and IT Engineering, Mazandaran University of Science and Technology, Babol, Iran 2- Faculty member of Electrical and Computer Engineering, Babol Noshirvani University of Technology, Babol, Iran 3- Faculty member of Computer and IT Engineering, Mazandaran University of Science and Technology, Babol, Iran

Keywords :

Iris Detection, Iris Segmentation, Convolution Neural Network, ResNet, Squeeze and Excitation Network

Abstract :

Iris segmentation is an important research topic that its objective is to obtain the iris area. There are many studies for iris segmentation. Most of the state-of-the-art iris segmentation methods are based on deep learning. These methods have improved the iris segmentation accuracy greatly. Now, the main challenge for researchers is to segment the noisy iris images captured in non-ideal environments created by visible light and user non-cooperation that received significant attention. In this paper, a new method based on deep learning is proposed for segmentation of the noisy iris images that includes two stages: detection and segmentation. In first stage, the iris region is detected using a Convolutional Neural Network (CNN) and an attention mask is created. In the second stage, the eye image is segmented using a new CNN, and the obtained attention mask is used to focus on the iris region. The proposed method also uses ResNet-18 in combination with Squeeze and Excitation (SE) block as the backbone in both stages. The proposed method has been tested on well-known iris datasets: UBIRIS and Iris Distance subset from CASIA, and has shown the good results in comparison with the other methods.