In this paper, a method is proposed for detecting and tracking eyelid movement from face image sequence for drowsiness detection of a driver. Conventional methods for measuring eye movement use electrodes, a heavy glass, a hard hat or an active light source, and these are not convenient for a driver. The authors propose to detect blinking by analysis of face image sequence. A monitor camera embedded in the dashboard does not disturb the driver. The two modes are initial mode and tracking mode. In the initial mode, eye region candidates are determined. Inner corners of left and right eye are first detected. Width of eye region is approximated by a distance between left and right inner corner of eye taking into consideration the fact that a distance between left and right inner corner of eyes are nearly the same as width of eye. The height of eye regions is decided empirically. In tracking mode, inner and outer corner of eye are detected. Then, each upper eyelid is approximated by a parabolic curve fitting and iris is detected as circle using simplified Hough transform. The proposed algorithm was applied to real face image sequences. The experimental results show the effectiveness of the proposed method as a tool for detecting driver drowsiness.