Can PIL or any other library deal with 16 bit greyscale or 16 bit/channel (36 bit or 48 bit) colour data? (Photoemulsions typically contain >12 bit signal, which is lost when converted to 8 bit data types).