This paper presents a comparison of different threshold-based selection methods, implemented and adapted for eye image segmentation. We used three subclasses of segmentation methods: fixed threshold segmentation (fixed threshold, quantitative threshold and cumulative distribution function method), global threshold segmentation (minimum error Kittler and characteristic separation), and local adaptive threshold (Bradley, Bernsen and Niblack methods). We compared the segmentation methods in order to determine their performances for video images and their suitability for eye tracking applications. In normal working situation using the eye tracking system we can use a global threshold value and in difficult conditions (different or fluctuating lighting) we can use a local adaptive threshold.

Keywords: Eye tracking, Image segmentation, Assistive technology, Threshold selection