Re: Media Testing for maclife.de
Re: Media Testing for maclife.de
- Subject: Re: Media Testing for maclife.de
- From: Uli Zappe <email@hidden>
- Date: Wed, 10 Sep 2008 17:20:14 +0200
Am 10.09.2008 um 13:46 schrieb Fleisher, Ken:
I think you are confused by what is being described to you. A single
ICC profile cannot account for multiple lighting conditions. In
other words, if a scene has more than one illuminant
"simultaneously" lighting the scene in varying amounts. What you
describe, and tested, are multiple illuminants as single sources in
the scene (not simultaneous). This is not the same thing.
If this was meant, then indeed I misunderstood it. Thanks for pointing
this out!
However, in this case I'm confused about what it could possibly mean
to "account for multiple lighting conditions".
For the simplicity of the argument, let's assume I have 2 illuminants
with exactly the 2 color temperatures that Adobe's two matrices are
built for, i.e. 6500 K and 2850 K.
I thought the point of Adobe's dual-matrix profile approach was to
compensate for multiple cases of single illuminants. In my simple
example, for an image lit by the 6500 K illuminant, it would use the
6500 K matrix, and for an image lit by the 2850 K illuminant, it would
use the 2850 K matrix (it would interpolate for illuminants in between).
Now let's assume I have an image of a cube one side of which is lit by
my 6500 K illuminant, another side by my 2850 K illuminant.
Are you really saying that an Adobe profile would be capable of
somehow applying its 6500 K matrix to the pixels that represent the
side of the cube lit by the 6500 K illuminant, while applying the 2850
K matrix to the pixels that represent the side of the cube lit by the
6500 K illuminant? I cannot imagine how this could possibly work. If,
on the other hand, the Adobe profile just takes into account the
average color temperature of both illuminants, then I fail to see the
difference to the situation with a "single simultaneous" illuminant
with something like 4675 K.
Another observation is that you are working on the assumption that a
raw
converter's default settings are the manufacturer's best attempt at
the
optimal settings. This an incorrect assumption for at least two
reasons.
1) A raw converter must be capable of interpreting images from many
different situations. There simply is no "best" setting. The default
settings are generally just null values for each of the options (or
sometimes a general guess like a sharpening setting). This in no way
implies
the appropriateness of the settings for any particular image. For the
"camera" profiles described that try to imitate the "look" of a
particular
camera, it is a little more accurate to assume the settings have
meaning,
but that meaning is in no way to imply number 2 (below).
2) The goal of all raw interpreters is to make pleasing images, not
color
accurate ones. Therefore, even if the default settings were meant to
create
the manufacturer's best attempt at the "correct" settings, the
target of
these settings is a pleasing image, not a color accurate one (as
Chris Cox
said about the Adobe products). So whether the CC24 patches hit their
targets or not is irrelevant.
I see your point, and this needs some clarifying. My premise
originally only applies to profiling software (this was the main focus
of the review, after all), not to RAW converters at all. True, you'll
need a RAW converter in an ICC camera profile workflow, too, but the
rule would always be to stick to exactly the RAW converter settings
that you used when building the profile, so for our discussion, we can
say the RAW converter as such is "neutralized". (Let's ignore any
deliberate modifications for aesthetic reasons *after* you got a
neutral image with your ICC profile.)
So if it's not the RAW converter that counts, but the ICC profile, I
think my premise is a reasonable assumption. You use an ICC profile to
get metrological neutrality, so every vendor of profiling software
will try to make sure that already with the default settings of the
profiling package, the ICC profiles created will approach this goal as
close as possible. (There are issues discussed earlier in the
maclife.de thread, when scanner profiling was discussed, namely
overfitting of curves for metrologically good results that have
problems with smoothness etc., so you can argue that metrologically
good results are not sufficient, but they are certainly necessary).
Default profiles of RAW converters were only tested in the Mac Life
review to see if ICC camera profiles could improve on them from a
neutrality POV, because obviously, if they can't, then there's no need
to use them. In other words, RAW convertor defaults were considered to
help readers decide whether they needed to invest in an ICC profiling
package or not. It's obvious, then, that the profiles of RAW
converters were evaluated from a neutrality POV, because otherwise,
you'd compare apples with oranges: of course there might be users of
RAW converters who don't want neutrality, but those would never
consider buying an ICC camera profiling package in the first place.
A second important point to make is that the review does not care
about a proof-of-concept or the theoretical potential of a specific
profile technology as such, only about what customers can actually
achieve with the different packages in real life.
So if I can build a neutral profile "out-of-the-box" using e.g.
ProfileMaker with a few mouse clicks, and could achieve the same
neutrality by manually tweaking some other profile for many hours, and
with a necessary background of years of experience, would you say
these two profile offerings are equally good offers for someone who
wants neutrality for his images? Obviously not, although you could
argue that "in theory", both profiles can "achieve" that.
With this in mind, let's compare 3 different products:
1. Canon's default profiles. Obviously, Canon goes for "look" mostly,
but the important point here is that they offer six profiles, not only
one. And one profile, "Faithful", is explicitly built for color
neutrality, according to the Canon manual. After confirming that
"Faithful" is indeed the most neutral profile of the six, I think it's
reasonable to assume that Canon tried to offer the most neutrality
they could with this profile.
2. A second case would be Aperture. Apple offers only one (completely
hidden) profile that users cannot change. Because of your point 2
above, you could argue it would be "unfair" to judge Apple according
to the neutrality of this setting, but the point remains that because
of this, Aperture is not a good choice for users who want neutrality.
Yes, maybe neutrality could be achieved by tweaking Aperture's
settings, but for someone who wants neutrality, that's not a good
alternative to a package that delivers neutrality out of the box.
3. Adobe is a more complicated case because it offers so many
alternatives (if you take the new beta profiles into account).
Obviously, you cannot blame Adobe at all for not improving on Canon's
original profiles with their emulations - after all, emulations are
all about identical behavior. As far as Adobe's "Standard" profile is
concerned, the situation is as with Aperture above; if anything, you
could ask why Adobe felt it was necessary to implement a proprietary
profile format to improve quality if quality in the sense of
neutrality isn't what they're aiming for, anyway. But - and this is
the important point - there is also the DNG Profile Editor that allows
to build profiles from the ground up by measuring a ColorChecker
chart. This is kind of a profiling approach built on metrology.
Therefore, I feel that it's justified to judge this technology
according to the metrological quality it can achieve out of the box,
without additional tweaking, cause that's exactly what it seems to be
built for.
Bye
Uli
________________________________________________________
Uli Zappe, Solmsstraße 5, D-65189 Wiesbaden, Germany
http://www.ritual.org
Fon: +49-700-ULIZAPPE
Fax: +49-700-ZAPPEFAX
________________________________________________________
_______________________________________________
Do not post admin requests to the list. They will be ignored.
Colorsync-users mailing list (email@hidden)
Help/Unsubscribe/Update your Subscription:
This email sent to email@hidden