Video-based eye gaze detection systems are useful for eye-slaved support systems for the severely disabled. The pupil center in the video image is a focal point to determine the eye gaze. Recently, to improve the disadvantages of traditional pupil detection methods, a pupil detection technique using two light sources (LEDs) and the image difference method was proposed. In addition, for users or subjects wearing corrective eyeglasses a method for eliminating the images of the light sources reflected in the glass lens was proposed. However, image-processing hardware for implementing these methods is rather expensive. In the present paper, the hardware construction is replaced by a construction consisting of a combination of a conventional image grabber and a personal computer. An algorithm for windowing around the pupil image with an automatic thresholding method for pupil detection is proposed. The results show that the algorithm works well when the user or the subject is wearing eyeglasses and under normal ambient lighting conditions. The calculation time is quick enough for real time processing. These algorithms would contribute to consistent and reliable pupil detection.