Basically puttig in a bogus (predetermined) out of range value on "the bad pixel" on to a "bad pixels record position" that will enable the processeor to detect it as a bad pixel and makes the image processor "fill-in" that bit based on its neighbouring pixels.