What's the fuzz about 16 bit?
See also my article about RAW mode. Basically, when you
are working on a digital image, what you are doing is making lots of
calculations - it may look like an image to you and it may look like you're,
say, increasing contrast, but to the computer it is just a big list of numbers
and some mathematical formula it has to throw at it.
If you have 8 bits per color channel, each channel can take any of 256 values.
This means that after every calculation, the computer rounds off to one of
these 256 values - it is not the case that you can have immediate values - 255
divided by 2 becomes 127 or 128, not 127.5.
Needless to say, the more calculations ("image corrections") you make, the
more rounding errors you accumulate and the farther the result will be from
the intended result. 255 by 2 by 2 times 3 will not become 191.25 but 189. And
you typically do lots of calculations...
With 16 bits per color channel, you simply have 65536 values to work with
instead of 256. 256 times as many values. 65535 divided by 2, again, times 3
becomes 49149 instead of 49151.25. The absolute error still is 2.25, but the
relative error is 256 times smaller so the visual end result will be
much closer to what you intended.
That's why it is important to stay in 16 bits as much as possible. Sadly, most
consumer level tools, as good as they may be (PaintShop Pro comes to mind), only will let
you work in 8 bits. Because I'm a sort of quality weirdo - always hunting for
it but never willing to spend the necessary cash - I found myself quickly on a
quest for affordable 16 bit software, and if possible 16 bit software that
knows a bit about color management.
The good news is: it exists, and you
will end up with something that costs a third of PhotoShop CS and will
probably deliver better results. Look here for the
details.