On 10/8/2010 7:08 AM, Ken Norton wrote:
> Moose spoke thusly
>> ... When scanning color neg film with an ICC profile ...
> Moose's ICC profile applied in the scanner software is really just applying a
> pre-programmed curve to the image file.
Sorry, gotta disagree here. Although the results appear similar in many/most
cases, the technology is fundamentally
different. And that can make a significant difference. As the film and cameras
we use have been rather well designed,
the amount of adjustment needed to get to accurate colors is generally modest,
so it's easy to assume that the two
approaches are essentially the same.
> It's a time saver for getting the image close to the way the film renders a
> standard scene in standard lighting to a standard result.
Again, not quite correct. The ICC color profile has no necessary connection to
film. The association is natural, as it
was developed at a time when film was the major medium for color imaging and
was and still is the medium for creation of
secondary (tertiary?) standards.
Nevertheless, an IT8 target could be created using colors created directly in
software and printed on an inkjet printer,
with no photographic emulsions involved at all. In fact, I would expect this to
replace the current technology sooner or
later. In theory, one could be created by painting, with ceramic tiles, etc.
An IT8 target is defined as a series of 228 color patches and 24 grayscale
patches, each with a defined value as
measured with a laboratory device (colorimeter, densitometer, whatever).
Because the creation process isn't accurate
enough, each target comes with a computer file showing the difference between
it's measured values and the correct values.
So, there are a couple of obvious differences between an ICC profile and the
curves that show film spectral response,
define the gamma of color image displays, are adjusted in editors, etc.
First, they define only a set of specific points. Intermediate values are
interpolated when software applies a profile.
This could be a problem if an imager were used that had sharp discontinuities
in its response graph. In practice, with
the technologies we use today, it's not an issue.
Second, they can define bumps and dips in response that no simple mathematical
formula can emulate. One may create
curves with funny wiggles and waves in editors, and I have occasionally done
so, but that's subjectively dependent on
one person's "eye" and usually not generically useful. And nothing like as
accurate as the defined results of an ICC
profile.
So what an ICC profile does is define the relationship between the color
response of any device, camera, film,
printer/ink/paper combo, CRT, LCD, etc., etc. and an idealized, virtual color
space. Thus, using the profiles of an
input and output device, an app may define the color mapping relationship
between them and 'perfect' colors. So when I
scan an image, then open it in an app using its profile and view it on a
monitor or printer using its profile, the
result is truer to the original subject colors than either device is capable of
on its own. The same is true of chains
of devices.
So what are the limitations of this system? I'm sure there are others, but the
primary one I'm aware of is that of
colors from iridescence, interference and diffraction phenomena, such as
rainbows, crystal refraction, newtons rings,
etc. Because the targets are reflective or transmissive, they don't provide
reference colors for different sorts of
light sources.
BTW, it is also possible, and easy, to scan without a profile, then later apply
the profile in PS - same result. This
can be a big help to anyone who has scanned a lot of film, then decides to
create/acquire and use an ICC profile. No
rescanning required. Also, one may apply a profile to a digital camera image in
post.
> Absolutely no different than AA's "Zone System" which was a calibrated way of
> doing ICC profiles before there ever were ICC profiles.
Welll, sorta. Except that ICC profiles come from lab measurements of
brightness, not a trained human vision system.
> Since Moose referenced VueScan, let's address that. When you do a scan with
> VueScan and a decent scanner, one can turn off all the color adjustments.
> Perform a preview scan and see how light or dark the image is. Altering the
> scanner exposure (first tab) is no different than adjusting the exposure in a
> camera--it's a way to get the basic exposure as close to YOUR raw file ideal
> as possible.
Yes, that works, using "Color Balance: None". Alternatively, one may use "Color
Balance: Neutral" and the black and
white point settings to spread the histogram precisely between the ends. At
least with my scanners, this gives a good
histogram using the full DR of the scanner with less work.
> Some with ETTR, some will EFTM, some with ETTL.
Once again, your love of jargon and acronyms have exceeded the limits of my
knowledge as to whatever those series of
letters may mean.
> The histogram is a great guide.
Yup, yup
> Once you have the entire histogram comfortably within limits
With the Neutral setting, right up to the limits, without going over, unless
you want to.
> (the Nikon scanners make this easy as they have gobs of DR), then you can
> scan and save this file making believe in your head that it's just a RAW file
> from a digital camera--as it is exactly that. The scanner is
> a digital camera and the scene just happens to be a macro picture of a flat
> see-through object, not a 3D distant image. So treat it as such.
Yup
> So, Moose's ICC profile is pretty much like the ICC profile digital cameras
> apply to the image data when you are doing in-camera JPEGs. Actually, it is
> no difference at all.
Nope, nope. See above.
> Ctein applies his own curves (which is all an ICC profile is) to the image
> file in PhotoShop. Moose applies them in VueScan. Same thing,
Nope. As above, the ICC profile is not a set of curves, but a set of defined
points of difference between the capture
color response and ideal color response. I usually separately apply Curves to
meet the vision I had when capturing the
image later in PS. With film, my experience is that I need much less adjustment
in post if I use a profile in scanning.
> the curves are applied in exactly the same way post-scan. The scanner is
> doing absolutely nothing different. Tabs 2, 3 and 4 in VueScan are all
> post-scan. Only Tab 1 addresses the physical traits of the scanner itself.
One nit. "Crop" does affect the area actually physically scanned, at least on
some scanners.
A. Profiled Moose
--
_________________________________________________________________
Options: http://lists.thomasclausen.net/mailman/listinfo/olympus
Archives: http://lists.thomasclausen.net/mailman/private/olympus/
Themed Olympus Photo Exhibition: http://www.tope.nl/
|