Re: Scanner Training Advice


Martin replies:

Sorry, I didn't make myself very clear on this point.

The light source on the flatbeds is extremely critical when it
comes to the
apparent sharpness of the images. IMO the first Scitex and
Scanview flatbeds
produced horrible, oversharpened images. The results from
prints were pretty much unusable unless the USM was turned

The smallest specs of dust or surface imperfections were
magnified, so your
high quality fibre print ended up looking a real mess.

Scanview have since modified the light source to make it less
focused and
more diffuse -- this cured the problem.
I don't think that things have changed, even on the top of the line
model purchased less than six months ago...

The main pre press guy who does most of the scanning comes
from a traditional background - USM in the scan.

As you and many others have noted, the EverSmart line have a
very crisp 'print' look - even with USM off. I have to agree.
Photographers who output to Lightjets or other contone devices
do not like this effect, and are quite vocal on it.

For transmissive scans, this seems an advantage - but I am
biased towards commercial flatsheet offset.

For transmissive, scaner USM may be OK, but if left to my own
devices I would sharpen reflective prints in Photoshop.

Reflective photo prints really suffer from the USM, they look
crunchy on the monitor - but it's amazing what an equalizer the
halftone screen is - even at 175 lpi using CTP.

The USM control in the scanner does have a good grain control,
plus there is also optical defocusing and then the red, green or
red/green filter combo in the scanner...

I think I will really have to pick the CreoScitex guys brain on the
extended USM functions.

Each time you check a new area of the image, its a new 'max
detail' scan. This can take about 30 seconds just to evaluate
This is similar to our scanner -- we get a general preview that
is fairly
mediocre -- but can choose to scan a portion at final resolution
committing the whole image.

Unfortunately you just have to keep a bunch of settings in your
mind --
ColorQuartet allows adjustment of strength, noise, light
contours, dark
contours and contour thresholds. With power comes the
responsibility to
Too true. Thats why it's so easy for many to stick to factory

There is a weird table called '' which I cant find any
mention of - but I presume it is something to do with

Lut is a look-up-table which is probably used to display your
image on
screen, or convert native Lab data to RGB.
I am familiar with LUT's, as this is our default method for
acquiring CMYK (or perhaps one day the odd RGB) scans.

What I mean is that you can choose two colour modes for the
scan. CMYK or perhaps RGB.

Then there are two real options for scanning into a specific
space or flavour, with a third option being the device link section.

The first option, is the input/output ICC profiles. You can scan
from ref/trans scanner RGB input to other output modes, or raw
scanner space (with or without a tag). This is all ICC driven.

The second option is to use the default RGB or CMYK table
option for separation or colour description. The input is the
default ref/trans RGB input description. This is the old style
format, proprietary CLUT approach. It is also used for the non
ICC Scitex monitor calibration/colour matching tool - since these
tables are out of the ICC softproofing link.

This 'table' section has the second option with the lable ''.

I recall reading somewhere online, but cant find where - that this
is used in ICC scanning or something...either IT8 scanning or
perhaps raw untoned high bit editing.

Either way, it probably is not a concern for us.

This provides a very flat, untoned image. But I guess it
has a use - I can sleep till I find out on the training day.

There are also 'device link profiles' which are a proprietary
combination of tables/profiles/scanner settings which are
presaved into a 'profile'.

This does not seem to be a true ICC profile - but a proprietary
Proprietory always bothers me -- which is why we don't buy
much Scitex kit.
This is true, up to a point - but I think times are changing.
Perhpas the Lino approach is giving them some competition.

The eversmart uses LS or luminance/saturation curves - but LAB
is not an option that I know of - even with ICC profiles. I dont have
a LAB ICC profile, only an old Kodak CM 'precision transform' file
which does not show up. So they are not directly going after the
Lino method of using LCH editing.

The proprietary issue for me is the default CMYK table - which is
fantastic for most work. This gives much better colour than the
ICC options - but not detail in some cases.

It is my guess that this is a gamut compression issue - or lack of
it. The default tables we like would use relcol rendering, so this
probably accounts for the lack of detail in some tough originals.

We rarely (never) scan in ICC mode, or as RGB and convert in
Photoshop - which does have certain benefits for certain

The 'Device Link Profiles' in question are proprietary - but only in
the sense that they are a scanner preset.

You can choose default tables or ICC methods describing your
RGB or CMYK scans - the device link profile is just a package for
holding these proprietary or ICC settings along with scanner
functions such as endpoints, gradation, USM etc.

The problem is that some of this is ICC, some is scanner
specific - all bundled up in a proprietary file. The confusion in my
mind is the word 'profile' - not all profiles are ICC profiles.
CreoScitex use the term, which creates confusion since the only
other option for defining RGB or CMYK is through ICC profiles,
with or wthout the use of these specific scanner settings.

One thing that I find amusing is how beauty is in the eye of the
beholder. I have read comments from noted ICC proponents on
other lists that they would not use the CreoScitex gear because
of how poorly it handles CM (or used to).

My problems are more petty.

If you do a full bed preview of many originals - before making
your various crops/scans you can only set the crops at this
postage size preview.

After you do an aptly named 'crop preview' - there is no way to
fine tune the crop, which must be done in Photoshop.

This way you are forced to overshoot, so that critical edge content
is not missed. But this is a workflow issue, the scanner is
designed for premounting of originals on registration pin
punched sheets - and the scanner automatically scans and
crops out the pre configured mask window positions.

If you do not have the image rotated correctly - there is no way to
do this in scanner software, only mirroring. You have to rotate the
image in Photoshop.

The prosumer Umax MagicScan software that I used to use did
all this and more. At least software can be upgraded, if
CreoScitex ever get a clue.

And one final observation which really amuses me - and is
probably why some pundits hate the scanner/software:

There is no histogram display (sorry to mention the H word Dan)

I find the Leaf approach of mapping a simple three point curve
over a histogram would make some levels users
more comfortable using curves.

The EverSmart softwares curve control has a nice feature which
is on my wish list for Photoshop - a diagonal reference line is
always dispayed.

This lets you quickly visually judge exactly how steep the curve is
in relation to neutral, as well as bringing the curve back to true in
the midtones and shadows, while you tweak the quartertones
(for example). Such a simple visual aid is very, very nice - and
seems obvious for Photoshop in hindsight.


Stephen Marsh.

Join to automatically receive all group messages.