Re: 16 bit versus 8bit
Re: 16 bit versus 8bit
- Subject: Re: 16 bit versus 8bit
- From: Lee Varis <email@hidden>
- Date: Wed, 06 Feb 2002 16:00:56 -0800
- Organization: Varis PhotoMedia
Bruce Fraser wrote:
>
Histograms are a lousy way to evaluate the efficacy of anything
>
unless you understand what you're looking at.
As usual, Bruce, you raise some very good points - especially about
scanning negatives, which, I have to admit I have very little experience
with because I never shot negatives in my work. I can see how, after
dealing with that orange mask you probably could use the 14 bits or so
of data in a good scan to optimize an image - very good point!
>
...The [16 bit] hstogram always
>
shows you exactly what you'd get if you downsampled the image to 16 [ 8?]
>
bits per channel. No shorthand involved.
If I understand you here (and I may not fully) this just points out that
the 16 bit histogram is just an optimized 8 bit histogram and I'm still
not sure that that is particularly useful except to demonstrate the
advantages of 16 bit over 8 bit in some "visible" way (it makes a good
screenshot for a book). Anyone else have some method or technique that
utilizes the info in the histogram?
>
>The other problem with 16 bit editing is that even if you work around
>
>the problem of making selections in a 16 bit file the selection is still
>
>8 bits and a gradient in the selection (like a feather radius of 60 or
>
>80) is still a severely "stepped" 256 level thing that can introduce
>
>problems all on its own...
>
>
This seems to contradict everything else you've said. If 8 bits is
>
enough for the data, why wouldn't it be enough for the mask?
Image data is generally distributed between 3 channels and unless a
gradient is neutral the "steps" in a gradient area won't line up - that
makes it harder to see because the values in the overlapping steps will
be closer together (except in LAB where "lightness" is in one channel).
Also, less noise is required to kill any color banding because, value
wise, its not as visible. A mask is another matter because you only have
one channel for your distribution of values. Another thing sometimes
happens when you have overlapping masks in multiple layers; you can get
a kind of interference between the frequency of opposing gradients that
can reinforce certain "steps" in the gradient. This can appear like bad
"blocky" jpeg artifacts or strange stairstep aliasing. The only solution
is to add more noise to the masks than was already there (by checking
dither in gradients options) You have to jump through a few hoops to do
this without introducing specks into the solid white or black areas of a
mask and its sometimes a PITA. At any rate, if you have a 16 bit file
wouldn't you want any selective corrections to blend in as smoothly as
the surrounding 16 bit data? I guess you would have to have worked a bit
with Live Picture composites to appreciate the difference and its not
something that would ever show up in anything but a large Lightjet print
but I have been bitten by this problem before. I also feel that the
otherwise excellent Extract feature would work a little better if the
mask rendered into the layer transparency could be calculated at a
higher bit depth for precision- you can still end up with some pretty
abrupt transitions.
>
...Then I can bring all that data into Photoshop, apply a profile that
>
represents the scanner's raw behavior, and start making those
>
critical decisions while being able to examine every pixel. It was
>
the crappy nature of most scanner software that led me down this path.
You are using an Imacon scanner mostly, right? This is actually the way
I work (some of the time) with my digital camera files (talk about
crappy software!). I may not see huge advantages to working with 16 bits
files but I am not above trying to squeeze every last bit of quality
from an image even if its just an imagined difference. Once I've worked
up an image I don't see the point of archiving 16 bit data though -
others may disagree. It's amazing to me that most of the supposedly high
end drum scanners don't actually allow you to save a 16 bit file. A
friend of mine (one of the original Quantel Paintbox engineers) had to
write his own software to intercept the data from his $150k ICG vertical
drum scanners so he could bring "untoned" 16 bit data into Photoshop.
The scanner software only saves 8 bit files after passing them through a
16 bit color correction stage and it cannot use or interpret ICC
profiles at all. It works with proprietary cluts for different film
stocks but its mechanical precision is awesome - cleanest, sharpest
scans I've ever seen.
>
If you don't need it, don't use it ... The trouble is, once you know you
>
needed it, if you don't have it, it's too late...
Another very good point! I have to admit I'm much more comfortable
knowing that I can do something with 16 bit files if I ever have the need!!
--
Regards,
Lee Varis
email@hidden
http://www.varis.com
888-964-0024
_______________________________________________
colorsync-users mailing list | email@hidden
Help/Unsubscribe/Archives:
http://www.lists.apple.com/mailman/listinfo/colorsync-users
Do not post admin requests to the list. They will be ignored.