32-bit support

I am having problem converting 32-bit depth into 8-bit depth. It seems that I can get the image's "silhouette" using the formula
value_8bit = value_32bit \* (qPow (2, 8) - 1) / (qPow(2, 32) - 1)

But if I multiply it again with (qPow (2,8) -1) or 255, the image is almost close, at least at "grayscale" mode. I can't find any references on the conversion.
