This paper presents a system that computes the head posture using video data captured by a webcam. With a user supplied bounding box of the facial area, the program generates color and gradient maps of the face. These maps are used in combination with other image processing techniques to detect the eyes, nose tip and lips of the user. Using novel methods of eye and lip region tracking, the eye corners and lip centroid points can be continuously tracked in real time. The coordinates of these points serve as inputs for computing the 3D posture of the face. The pose parameters are next Kalman filtered and transmitted over the Internet where it will animate an avatar head to mimic the remote user. The program has been tested on a variety of subjects and is robust in facial feature detection as well as in real-time tracking. The resulting avatar movement closely mimics the movement of its real life counterpart. A framerate of 9fps can be achieved on a Pentium III 600.