This paper presents a novel approach to precisely locate eye position in near-infrared facial images. In this approach, we first determine the face region and initial eye position using face detection classifier based on Haar features and AdaBoost algorithm. Then we detect the eye edge in the eye region using Sobel operator, fit it into an elliptical contour. Finally, the center point of eye is located by the center of the fitted ellipse. With 120120 normalized face images, the experiments show that the proposed approach is accurate. The average error is less than 1.5 pixels and the processing time is about 7 ms.