Modeling Facial Skin Motion Properties in Video and Its Application to Matching Faces across Expressions