ISSN 1000-1239 CN 11-1777/TP

Journal of Computer Research and Development ›› 2020, Vol. 57 ›› Issue (2): 395-412.doi: 10.7544/issn1000-1239.2020.20190092

Previous Articles     Next Articles

A Benchmark for Iris Segmentation

Wang Caiyong and Sun Zhenan   

  1. (School of Artificial Intelligence, University of Chinese Academy of Sciences, Beijing 100049) (National Laboratory of Pattern Recognition (Institute of Automation, Chinese Academy of Sciences), Beijing 100190)
  • Online:2020-02-01
  • Supported by: 
    This work was supported by the National Natural Science Foundation of China (U1836217, 61427811) and the National Key Research and Development Program of China (2017YFC0821602).

Abstract: Iris recognition has been considered as one of the most stable and reliable biometric identification technologies. In the whole process of iris recognition, iris segmentation is in the preprocessing stage, so the quality of iris segmentation can directly affect the accuracy of iris recognition. Since Daugman first proposed a high-performance iris recognition system in 1993, a variety of iris segmentation algorithms have been proposed. Especially in recent years, deep learning based iris segmentation algorithms have greatly improved the accuracy of iris segmentation. However, due to the lack of unified databases and evaluation protocols, the comparisons of different iris segmentation algorithms are messy and absence of fairness, hence an open iris segmentation evaluation benchmark is proposed. First, the definition and challenges of iris segmentation are briefly introduced. Second, a comprehensive summary of three representative, open iris segmentation databases including the characteristics and challenges is given. Next, evaluation protocols of iris segmentation are defined. Then the traditional iris segmentation algorithms and deep learning based iris segmentation algorithms are elaborated on and compared by extensive experiments. The experimental results show that deep learning based iris segmentation methods outperform the traditional approaches in terms of accuracy. Finally, we make in-depth discussions about open questions of deep learning based iris segmentation algorithms.

Key words: biometric identification, iris recognition, iris segmentation, deep learning, semantic segmentation

CLC Number: