When the user initially clicks the eye feature, a box is drawn around the square and the subimage within this square is cropped out of the image frame. The cropped image is used as a template to find the position of the feature in subsequent frames.