It is not so simple. The characters are not the same colour as in the image. Look at the eyes and the nose for example. It seems to be doing a more advanced kind of image feature detection.