Abstarct: Eye detection and eye tracking techniques plays important roles in developing Human Computer Interaction (HCI) systems. Disabled aid applications, Driver fatigue detection and human identification are some examples. Efficiency of such systems is evaluated with respect to their accuracy in detecting and tracking of eyes in a video stream. The most important problems in detecting and tracking of eyes are head position i.e. rotation and orientation, obscure eyes due to glasses, hear and severe illumination conditions, etc. Therefore, many researches are being conducted to increase the reliability of these systems. In the first part of this thesis we are concerned with developing a robust method for detection of eyes in face image. The proposed method provides a solution which is invariant against similarity transformation and also robust to severe illumination conditions. Instead of directly using intensity information for describing the eyes in the face image, we represent the image using topographic labels. Using a set of functions invariant to similarity transformation, we describe the labeled topographic image. Finally baxsed on the extracted features a Bayiasian classifier is designed to detect eyes in the face image. In second part of the thesis, an eye tracking technique for video sequences is introduced. In order to track the eyes in the next frxame, we search for a region at the neighborhood of each eye in current frxame for which the correlation function is maximized. Thanks to the correlation function, the computational complexity of the method is considerably low. In the last part of research, a blxink detection method is proposed, we simply measure the distance between eyelids for this purpose by mapping of gradient image of eyes in vertical direction. The detection results applied on “XM2VTS” databaxse confirms the superior performance in different imaging conditions. The result of eye tracking using the proposed method was very promising.