Re: fine art reproduction questions
Re: fine art reproduction questions
- Subject: Re: fine art reproduction questions
- From: Klaus Karcher <email@hidden>
- Date: Fri, 07 May 2010 20:08:26 +0200
Dear Stanley,
Sorry for the late reply -- I'm a little bit busy lately (e.g. with
traveling around and profiling Cruse scanners ;-) My statements might be
biased as I'm working as freelancer for Cruse.
I'm sorry to hear of your rather disappointing experience with your
Cruse Scanner and want to get to the bottom of your issues.
Stanley Smith wrote:
Pasting a reply to a similar query made last week:
We have had our Cruse Scanner for several years now, and overall I
would say that it is a bit of a disappointment. We did work with
Cruse to fix a major flaw in the original machine, which was an
inability to achieve color consistency-- due to the lights reflecting
off the painting back into the opposite light fixture-- thereby
changing the color of the light hitting the painting in a
continuously variable way as the painting traveled under the lights
during the scan. The solution was to install special florescent
tubes called aperture tubes-- they emit all of their light through a
narrow slit in the otherwise opaque tube-- thus eliminating the need
for a reflective-backed light fixture. I identified a lighting
company in Hollywood (KinoFlo) who worked with Cruse to develop these
tubes, and I had thought that they would become standard. The fact
that apparently Curse still sells the original configuration, and has
not switched to the aperture tubes as a standard is a little
shocking.
Light reflected back from the lamp fixtures is definitely an issue.
However I'm skeptical about the usefulness of aperture tubes: They turn
a basically flat light source into several zoned strips. That's no
problem with flat, matte originals, but when it comes to specular,
uneven or 3-dimensional objects, the striped lighting can lead to
unattractive reflections and shadows. There are several other
disadvantages of aperture tubes, e.g. the more directional light
requires additional actuators to angle the lamps optimally for each of
the various scan modes, aperture tubes have a weak light output
resulting in long scan times and bad signal to noise ratios, ...
Alternative light sources are in development, but there is also a way to
improve the situation significantly with conventional fluorescent tubes:
I developed a method to compensate for variations caused by
backscattered light and other sources of error. Even though my solution
it not implemented into the scanner software yet, two very experienced
and reputable clients use a preliminary implementation for some time now
with great success: Art Petrus <http://www.artpetrus.com/> and LUP AG
<http://www.lup-ag.com/>.
I don't know if Peter Radzim (Art Petrus) or Elmar Lutz (LUP) are
lurking colorsync-users and like to jump in. Ohterwise I'd be pleased to
put you in touch with them off-list.
Other problems:
-- Since focus is automatic, and dependent on measuring the height of
the original from the scanner bed, it seems to miss more often than
hit-- resulting in a "trial and error" approach.
-- I would say that even though you can end up with a very high pixel
count the images are just not as sharp as you would expect.
hmm... I've never seen a Cruse scanner that needed trial an error to get
the focus right. Simply enter the height of the original into the
appropriate text filed of the scan software and you should be done. You
don't even have to measure the height very precisely: the depth of field
is at least several millimeters up to several centimeters (depending on
the lens and reproduction scale). It might happen that one enters the
height into the wrong text filed by mistake or forgets to update the
height for the next scan, but if you need to focus by trial and error or
get bad sharpness in general, you should commission Cruse for service.
-- The proprietary nature of the software makes it impossible to
create good custom profile-- you do not have access to a "raw" file--
just a rendered 16-bit tif. Thne result of this is a lot of post
processing to get the colors right.
This 16-bit TIF is almost the "rawest" image file you can get from a
trilinear CCD sensor. In contrast to CFA sensors, you get three
full-fledged line images per cycle out of the sensor and the image
processing is very simple -- there's not much "magic" in the software:
It is just about offset- and gain adjustment, black- and white
equalization and (optional, adjustable) gamma correction. It makes
absolutely no sense to start profiling with anything "rawer" than this.
It's definitely /not/ the proprietary software that makes it hard to
create good profile -- the difficulties arise from the spectral
properties of the originals, the peaky spectrum of the light source and
-- most of all -- the fact that no CCD fullfills the Luther Condition.
I'm engaged in getting the best out of Cruse scanners for more than 2
1/2 years now. I'm responsible fore some hardware improvements, have
developed a prototype solution to compensate for original dependent
variations, but most of the time I've spent with optimizing the profile
creation and developing special profiling targets (I'm also involved in
a research project on scanner targets for historical paintings).
I've outlined my profiling process in this post
<http://lists.apple.com/archives/colorsync-users/201/Jan//msg00054.html>
The overall results are amazing:
- average Delta E 2000 reduced by 70% to 80%
- max. Delta E 2000 reduced by considerably more than 50%
compared to conventionally profiled and configured scanners.
Even with the most demanding test sets (contains pigments that are
extremely prone to metamerism errors) I get an average Delta E 2000
about 1.5 (instead of approx. 4-6 with conventionally profiled and
configured scanners)
In well-defined areas of application I can even get average DE00's
significantly below 1!
Profiling in this way is definitely nothing one can accomplish with
off-the-shelf profiling applications and targets. I would definitely
*not* expect of the end customer to perform this torturous process ;-)
... and it's also not necessary for him to buy umpteen (expensive),
custom-built profiling targets and to familiarize with command-line
driven profiling tools and statistical computing environments: Once the
system is profiled at the final location, there is no need to re-profile
it unless one changes mayor parts like the CCD, IR filter or uses a
different tube type. Everything else can be well compensated with the
usual calibration tools and my software.
But the effort pays off: In my experience, accurate Argyll profiles
combined with measures to reduce scan-to-variability can even outperform
spectrally-aided workflows like Ergosoft's HP Artist solution for Nikon
or Betterlight's ColorSage. (It's not just my opinion: a German company
with many years of experience in fine art reproduction compared both
systems came to the same conclusion.)
I guess even most multispectral imaging systems will not be able to
achieve this /colorimetric/ precision (as Berns pointed out there is a
substantial difference between spectral an colorimetric accuracy: a
small spectral RMSE is not necessarily a equivalent of a small
colorimetric error!)
-- It is slow.
It typically takes less than 10 minutes to scan one square meter at
300dpi. How long does it take to capture and stitch the same with a DSLR
or MF digital back? How long to set up light and camera? Do you get the
same quality? (I'm talking about 3x260 million /real/ pixels per square
meter -- no demosaicing, no interpolation).
You can switch between different lighting setups and texture modes with
a single mouse click -- there's no need to setup or adjust anything. You
can even make batch-scans with severeal modes if you are unsure which
one best suits your original.
-- it is a very small company with limited resources and questionable
longevity.
You are concerned about the longevity of the /company/? I can relieve
your: Cruse is a small, but rock solid company with world-wide
reputation and customer base and ambitious development targets.
-- it is a VERY expensive machine, that takes up a lot of floor
space.
well: extraordinary demands require extraordinary solutions :-)
-- with current stitching software, it is becoming less necessary to
capture a very high rez file in one capture.
-- the "gang-up" approach proved to be difficult to set up. It
turned out to be faster to capture smaller objects on a traditional
copy stand one at a time.
The machine does have its positive aspects-- when it works and is
properly calibrated for exposure and focus, it does produce
spectacular files, and is especially good for large maps. However,
it is really slow, and a bit difficult to use. These days we use it
only occasionally, relying on a Betterlight set-up for most high-rez
work.
I don't see any advantages in Beterlight over Cruse except for portability.
After having talked to Cruse, I can confirm that you are invited to test
the advanced Scanner configuration and profile capabilities as well as
my (yet experimental) linearization tool.
Pleas contact Mike Lind, Cruse or me to arrange further details if you
are interested.
Regards,
Klaus Karcher
--
Klaus Karcher * Eichenallee 18
26203 Wardenburg * Germany
Tel.: +49 441 8859770
Mobil: +49 174 7283060
_______________________________________________
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