Today's facial-recognition systems use a camera to zoom in on an individual's eyes, mouth and nose, and extract a "feature vector, " a set of numbers that describes key aspects of the image, such as the precise distance between the eyes. (Remember, in the end, digital images are just numbers.) Those numbers can be fed back into a large database of faces in search of a match.
WSJ: Eric Schmidt and Jared Cohen: The Dark Side of the Digital Revolution