Re: 8 bits versus 16 bits edit
Re: 8 bits versus 16 bits edit
- Subject: Re: 8 bits versus 16 bits edit
- From: Stephen Marsh <email@hidden>
- Date: Tue, 08 Jan 2002 11:45:22 +1100
Paul writes:
>
> I edit most of my images in 8 bits with a curves layer. I have thought of something and wish to ask you if this method can
make a difference:
1) From the 8 bits file with curve layer: Save the curve, then trash it.
2) Convert the file in 16 bits, then apply the curve that has been
saved,
3) Finally revert to 8 bits and save the file. Is this supposed to
produce a smoother result than simply flatten the 8 bits image in the
first place? <<
Paul, there are two common scanning workflows:
* Edit in scanner for as many edits as possible, this will be performed
in high bit mode taking advantage of hardware and software in the
scanner where possible. After internal processing from higher bits per
channel, a 8 bpc image is the result. Some software allows the high bit
file to be saved, instead of reducing it down to a regular bit level.
* Lock down settings or 'dumb down' the scanner with fixed settings and
scan every file in high bit mode with the correct custom scanner
profile. An image editor is used in lieu of scanner software to tone and
colour correct the file in high bit mode. This file or a dupe can then
be reduced to 8 bpc mode for a more regular workflow and output.
This second approach seems to be what you are doing - but in 8 bpc mode?
If so, this is not ideal since you are working on lesser data than if
you were accessing the scanners native hardware/software. Dumbing down
the scanner and not editing in scanner is 'limiting' by definition - so
this is where high bit workflows become popular, since an image
processor is now performing the scanner related tonal/colour tasks.
Now for your questions:
Changing a file which starts in 8 bpc mode to 16 bpc for tonal editing
will open up more _values_ for the tonal processing to work with.
However, it is my guess that this is not often done, although there are
times when it is done, as in gradient banding reduction which can
benefit from this technique (due to the dither added from 16>8 bit mode
change).
The preferred workflow is to _start_ and edit in high bit mode for as
long as possible and then reduce to regular bit mode for further edits
or output. There is some debate in some circles on whether it is worth
bothering with high bits for some workflows and output. As in many of
these things, only the end user can make the decision on results vs.
time and other factors such as archival use etc.
In my high bit workflow tests, the only visual difference betweem the
two was edits in high bit looked as if it had a 0.2 pix gaussian blur
applied (it looked almost imperceptably softer on a hi res file). This
was monitor, on output there was no apparent difference (halftone and
stochastic output tend to average out small detail, although contone
output may differ).
Try the this URL for further reading:
http://www.ledet.com/margulis/ACT_postings/ACT-8-bit-16-bit.html
>
> Now, the same for converting into another profile: 8 bits or 16 bits? <<
The end user does not have to worry about trying to make the colour
transform more precise in Photoshop by changing bit depth (they are
independent). Mode conversions in ver5 were done in 16 bit for tables
and 20 bit for profiles...as APS6 only uses profiles, I would guess that
it uses 20 bit for all colour transforms - even when working on 8 bpc
data.
http://www.ledet.com/margulis/ACT_postings/ACT-Banding-16 bit-8-bit.html
As the two links above demonstrate, the Applied Color Theory list has
kicked this over many times before, as I suspect this list has (perhaps
try the CS archive).
Hope this helps,
Stephen Marsh.