Automatic threshold-setting method for iris detection for brown eyes in an eye–gaze interface system with a visible light camera

• An automatic threshold-setting method for iris detection is proposed.
• Eye images are analyzed in terms of saturation and intensity.
• The optimum threshold value for the saturation is obtained by discriminant analysis.
• The intensity threshold is determined based on the saturation threshold results.
• The method reduces the time required for threshold setting with reasonable accuracy.

This study describes the improvement of an eye–gaze interface system with a visible light camera. The current system detects the center of the iris from a captured eye image using image processing. During the initial stages of system use, a display window is provided to set the threshold values of the image׳s saturation and intensity, which is used to manually adjust the appearance of the iris region. In this study, we propose an automatic threshold setting method. The optimum threshold value for the saturation is obtained by discriminant analysis and that for the intensity is determined by finding the value that yields the same number of accumulated pixels in the detected region as threshold processing of the saturation. In our experiments with subjects with brown eyes, the automatic method obtained good threshold values in most cases. Furthermore, an adjustment function to overcome under- or over-estimated saturation threshold values is also proposed. This function provides a more robust automatic threshold setting. In experiments, we compared our automatic setting method with conventional manual techniques, which showed that the automatic method is useful for reducing the time required for threshold setting and its pointing accuracy is comparable to that of the manual approach.