Accurate Iris Center Localization Method Using Facial Landmark, Snakuscule, Circle Fitting and Binary Connected Component
Feng Xiao,Kejie Huang,Yue Qiu,Haibin Shen
DOI: https://doi.org/10.1007/s11042-018-5787-x
IF: 2.577
2018-01-01
Multimedia Tools and Applications
Abstract:Iris centers have been widely used in machine vision for face matching, gaze estimation, etc. However, in low resolution eye images, the iris and its surrounding region present a variety of appearance characteristics, which make it difficult to accurately locate the iris center. In this paper, we propose a robust, accurate and real-time iris center localization method by combining the facial landmark, snakuscule, circle fitting and binary connected component. Facial landmarks are used to extract an accurate eye Region of Interest (ROI). Thereafter, a fixed size circle-based active contour snakuscule is used to detect the iris center. Based on the snakuscule center and inner radius, a novel method is proposed to extract accurate iris edges for circle fitting. In addition, the quality of the detected iris center is evaluated by a circle-binary quality evaluation method. Binary connected component method is used to improve the accuracies in those unqualified images. The proposed method is tested on three publicly available databases BioID, GI4E and Talking Face Video. The result shows that it could achieve an accuracy of 94.35% on the BioID database when the normalized error is smaller than 0.05, which outperforms all state-of-the-art methods.