Re: OT: 16 bit editing myth or reality?
Some clarification if I may, in my first message on this topic, I tried to
write about the experiment I did and that it was to find out how many levels
where lost by applying a gamma correction on both 8 bit and 16 bit data. The
16 bit data was converted back to 8 bit integer after the application of the
gamma correction for comparison with the 8 bit data on equal basis. After a
few replies, I realised that what I did had nothing to do with compairing 8
and 16 bit editing because I compared each data set with itself, before and
after the gamma correction and then put the result of each side by side. I
tried my best to say this a few times but it seems I failed, hopefully this
time will be the good one.
In my last two messages I suggested two different approach to evaluate the
difference between editing in 8 and 16 bit mode by comparing two real image
with each other, first using the Lab colorspace which seems a bit
complicated for me to explain correctly and I probably made a mess of it.
Then tried again in my last message to use a more visual and direct method
of comparing two image by simply substracting one from the other plus an
offset (not to loose negative values if any). Then I basically said that 16
bit editing provides no visible benefit, this is assuming the higher bit
image is converted to 8 bit for the comparison and that the exact same
editing as been applied to both image, this exclude all kind of random
transforms. Here is why, I actually did this for both a 16 bit B&W image
and a 48 bit color image and in both case after Apply Image -> substract
offset 128, the resulting image was a visually uniform mid gray level (128).
In the case of the B&W image I couldn't see variation in the levels until I
magnified it at 100% and at 200% for the color image. I have to admit I'm
relatively old and my vision is probably not what it was when I was a
teenager. The histogram of both the B&W and color image where almost a spike
but not quite at the 128 level. All this as to be put in context, though the
histogram showed there was some relatively small but visible variation in
levels, when looking at the actual images side by side, these small
difference practically vanish or become invisible. If I printed all of them
(4) and showed them to a croud of a 100 people and asking them to pick the
image that was edited in 16 bit, I would bet you a large sum of money that
the result would be equivalent to the result of tossing a coin to choose the
16 bit image.
As I said earlier, this won't change my usual workflow, I'll still do 16 bit
editing and I didn't do this to change peoples mind about 16 bit editing I'm
just curious, some would say I have time to waste and it's true.