Re: 8 bit vs 16 bit workflow
Re: 8 bit vs 16 bit workflow
- Subject: Re: 8 bit vs 16 bit workflow
- From: neil snape <email@hidden>
- Date: Fri, 03 Aug 2001 11:44:04 +0200
on 3/08/01 10:37, Martin [apple account] at email@hidden wrote:
>
The debate on 8 bit vs 16 bit colour is hotting up on Dan Margulis's Color
>
Theory mailing list.
>
Dan's position is that in "real life" retouching there are no visible
>
benefits to working with 16 bit images.
Dan has more tricks than a magician at a big name circus of 'prepess'. He
has great info and tips and can tame the ugliest beast into manageable
images. Shall he ever get up on the podium at Speaker's corner...
Always inviting the chance for reflection, he likes to state things as they
are. High bit output devices don't exist so unlikely to see an improvement
on the output side. Image output devices are not foreseen to be in high bit
either for obvious technical reasons. Therefore yes 16bit files have little
or no visual difference on output.
>
And further, that if you take an original image in 16 bit, duplicate it,
>
dump it to 8 bit and carry out a whole bunch of complex corrections to both,
>
the final images are almost completely indistinguishable. The only
>
difference usually being an apparent softness in the 16 bit image (similar
>
to the softness you see when you compare a 100Mb scan with a 150Mb scan of
>
the same image).
This is where pure math vs. what you see will bring this discussion up
front. Since all transforms pass through numeric lut's there will be
rounding errors. One could say insignificant but look closely and you'll
easily see that 16 bit keeps many more levels in check (when converting back
to 8 bit for output) than heavy edits in 8bit. Bruce Fraser wrote much ado
about this in his book and for creativepro.com. The difference is as you say
like an image which has added detail to fill in the gaps. Now reality, as
you asked for from someone that does imagery rather than talk about it. Yes
there is an obvious advantage to 16 bit especially negative scans or digital
input, as you said when there is heavy editing to do. Yes you can see the
results in the file( voir measure by number). Output should be almost the
same though depending on the accuracy of the rip and stability of the
device. An imagesetter should be able to do graduates from a correct file
without banding whereas if the file has banding from heaving editing it
should produce this banding depending on screen angles etc. That's why this
subject really is intangibly arguable.
>
This year marked a decade of scanning and retouching for me. In those ten
>
years I've seen a handful of images that have left me thinking that I should
>
have been working in 16 bit (most of them scanned from colour negative).
Much longer than I, yet we'll agree on the above observation.
>
Lots of people here advocate 16 bit workflow, but can anybody provide
>
conclusive "real world" proof of the benefits of 16 bit acquisition and
>
manipulation? (The only caveat I suggest is that the images supplied as
>
proof should be "real" not some computer generated test target).
You can see differences as I said earlier but since you and I both think
that it's not needed often and assume when it is we apply 16bit, then we'd
have to print out both 8 and a 16 bit image to press. On a web press good
luck , then we'd need to go to a 175 or 200line offset, which as much as I'd
love to do... Lightjet 2080 film recorder can and does see finer graduations
though if the file has abrupt channel values falling out.
>
My own experiments tend to support Dan's case (comparing 8 bit and 14 bit
>
scans from drum, CCD and Volare camera).
Yes..
>
The majority of our images are scanned between 75 and 200 megs (RGB). If we
>
seriously look at doubling this data and losing 85 per cent of Photoshop's
>
capabilities there had better be some good reasons to do so!
The larger images have more detail spread out over the area to maintain so
in any case there will still be enough data to print. The dithering or
interpretation by the rip making a median of the pixel values will no doubt
smooth out differences anyway.
Neil Snape email@hidden
http://mapage.noos.fr/nsnape