Unless I'm mistaking the question, because the data need to be normalized later, you need per pixel intensity in the raw data.