An Algorithm for Eye Detection and Tracking in Face Images

To detect and track eye images, distinctive features of user eye are used. Generally, an eye-tracking and detection system can be divided into four steps: Face detection, eye region detection, pupil detection and eye tracking. To find the position of pupil, first, face region must be separated from the rest of the image using mixture of Gaussian, this will cause the images background to be non effective in our next steps. We used the bag of pixels technique from face region, to separate a region containing eyes and eyebrow. This will result in decreasing the computational complexity and ignoring some factors such as bread. Color entropy in the eye region is used to detect pupil. In the next step, we perform eye tracking. We proposed algorithm for the eye tracking by combining the pupil based Kalman filter with the mean shift algorithm. After locating the eyes in the initial frames, the Kalman filtering is activated to track a region containing eyes and eyebrow. If it fails in a frame, eye tracking based on mean shift will take over. In the proposed method, eye detection and tracking are applied on testing sets, gathered from different images of face data. Experiments indicate correct detection rate of 93.8%, which is indicative of the method’s superiority and high robustness.