To meet the real-time requirement of eye-controlled human-computer interaction systems, a fast and non-intrusive eye-tracking method was proposed. Firstly, CamShift algorithm and AdaBoost algorithm were combined to extract eye regions. Secondly, iris was located from the eye region in the first video frame. And the iris region was used as a template. Finally, the iris template was scaled according to new size of eye region in the subsequent video frames. Then the scaled template was used for matching in the eye region. The relative displacement between iris center and canthus or nostril can be used to control the movement of mouse pointer on computer screen. The experiments on a video of resolution at 640*480 result a tracking speed of 32 frames per second, and the correct rate is 95.5%. Results prove the effectiveness and efficiency of this method. So, this method can be directly applied to eye-controlled human-computer interaction systems.