Is there a reason whyt denoise algorithm don't work on 32bit floatting point image

can I assume that you do dark-frame subtraction and similar operation with a “flat field”?