Affiliations: [a] State Key Laboratory of Robotics, Shenyang Institute of Automation, Chinese Academy of Sciences, Shenyang, China
| [b]
Institutes for Robotics and Intelligent Manufacturing, Chinese Academy of Sciences, Shenyang, China
| [c] University of Chinese Academy of Sciences, Beijing, China
| [d] School of Computing, University of Portsmouth, Portsmouth, UK
| [e] Space Technology for Autonomous & Robotic systems Laboratory (STAR LAB), Surrey Space Centre, University of Surrey, Guildford, UK
Abstract: Gaze tracking has wide applications such as in driver fatigue detection, virtual reality, and human-computer interaction. The performance of gaze tracking depends largely on the accuracy of iris center localization. However, most of the existing gaze tracking products are intrusive or require additional equipment with a high cost. Therefore, precise localization methods of iris center in low quality images captured in a non-contact way with visible light need to be investigated. This paper proposes a novel localization method of iris center using energy map synthesis based on image gradient, isophote and midpoint of eye ROI (Region of interest). This method combines the advantages of higher localization accuracy based on gradient, invariance to the rotation and linear transformation of light based on isophote, and iris center close to the midpoint of eye ROI. Moreover, a post-processing correction method for the closed eyes and for other large deviations of iris center position is adopted to further improve the localization accuracy. The algorithm is verified on the BioID, Talking Face Video and MUCT Face databases, and the results show that the localization accuracy in the paper has outperformed the listed state-of-the-art methods in varying illuminations.
Keywords: Iris center localization, energy map synthesis, gradient, isophote, post-processing correction
