You'd expect so, because there are already projects which do sub-pixel image recovery from video frame. So putting that before recognition should be doable.