cheapie color profile solution? (Epson 4870)

[false_dmitrii]

Hi again,

When I scan color photographs on my Epson 4870, many colors come in
properly. However, reds (& pink, orange, brown, etc) typically shift
noticeably. After some careful evaluation, I've reached the tentative
decision that most of the discrepancies are due to the fluorescent
light source. When I held my test photo near the light source, the
colors were closer to the scanner's onscreen output than their
appearance under normal viewing conditions. I was unable to correct
the red via PSP8 or PSE2--it was too complicated a problem for my
knowledge level--and while I'm sure there's a way I'm not aware of,
I'd prefer to solve the problem at the source via color profiling.

My scanning is for personal use with an eye toward long-term viability
of the initial files. I'm willing to put some extra money toward the
scans themselves that I can't justify for image editors, printing,
etc. I don't want to spend hundreds on a colorimeter if possible.
With that in mind, what sort of color management should I be looking
for?

Specifically, is buying an IT8 target and profiling the scanner
through Vuescan or freeware truly useless without an accurately
calibrated monitor? I've never done a profile and don't know exactly
what's involved. I'm using a default manufacturer profile right now
and am satisfied with its output. Good digital camera photos display
with colors that appear to be correct; if I can match their color
quality with the scanner I'll be happy. Is there any reason ICC
profiling the scanner would make things worse rather than better?

Even if the monitor is uncalibrated, does it make a difference to the
initial scanner output? Does a user-created scanner ICC profile
depend on his monitor profile for accuracy, or does an inaccurate
monitor profile merely prevent him from making accurate postscan
adjustments to the original scan output? In other words, if the
original profiled scan output is viewed later on a profiled monitor,
will everything look right?

Otherwise, does it sound like I'm doing something else wrong with my
scans to cause the problem?

Thanks in advance,
false_dmitrii

 

[Douglas MacDonald]

Are you sure the colour shift is not you monitor? I have the Epson scanner
and a monitor calibrated monthly. The scanner - out of the box, has perfect
colour on just about every negative I've fed it.

Douglas

 

[false_dmitrii]

Are you sure the colour shift is not you monitor? I have the Epson scanner
and a monitor calibrated monthly. The scanner - out of the box, has perfect
colour on just about every negative I've fed it.

Douglas

Unfortunately, I'm not sure. And I'm stuck on an LCD screen until I
can upgrade--soon, I hope--which introduces other limitations. Any
suggestions on how to do a passable eyeball test of the monitor's
accuracy without professional color profiling tools?

The photo I was testing was a 4x6" color print. I'm suspicious of the
scanner because the color shift seemed close to what I saw when I
turned off my other lights and held the test photo next to the
scanner's glowing fluorescent lamp. In other words, the scanner may
be capturing a near-perfect image of a wrongly (for my tastes)
illuminated print. If this is the case, I'd like to correct it toward
incandescent lighting. I fiddled with PSP8's Automatic Color Balance,
which offers color temperature guidelines, but as the reds improved
other colors started drifting. I tried PSE2's color replacer without
much luck...or experience.

What's your method for removing negative media color cast? I'm still
playing with Epson Scan, and its thumbnail mode doesn't seem well
suited for setting a black point from the blank lead-in of the film
roll or tracking one down in the non-zoomable preview window. Instead,
I've been manually adjusting the histogram to trim out the blank areas
at the low and high ends of each channel with minimal clipping. I
then usually boost saturation to 50. It's not perfect, but it seems
to produce better results than the auto exposure. To my eyes, the
adjusted scans are close in color to their corresponding third-party
prints, as well as whatever real-world objects I can use for
comparison. They're slightly off, but not as much as my scans of
opaque media. For these color drifts my manual adjustments could be
at fault.

Any other thoughts?

false_dmitrii

 

[false_dmitrii]

Are you sure the colour shift is not you monitor? I have the Epson scanner
and a monitor calibrated monthly. The scanner - out of the box, has perfect
colour on just about every negative I've fed it.

I should also mention that I have Epson Scan set to "ICM", "Source:
Epson (Standard)", and "Target: sRGB". Perhaps one of these is the
wrong choice?

false_dmitrii

 

[nikita]

If you just want to bump that LCD into a decent ballpark without using
hardware calibration tools, there is just one eyeballed solution to
this. If you give it some time and do it very carefully, you could
take it quite close. The Supercal adjusts the curves in many fine
steps from black to white in each channel. This is MUCH better even on
a CRT and Adobe gamma. The possibility to calibrate and profile a LCD
by eyeballing is quite unique.

It's free. But you should pay that little fee they're asking for.

You shoulden't need to tweak that Epsonscanner that much.
But try Vuescan if you do already have an IT8 target to scan and use.
Also, if not having the IT8, just try different workingspaces in
Vuescan even if it's not the same as in Photoshop. You could see that
as different "premade profiles", as Vuescans custom profiling is built
upon a very simple profiling base. If you're lucky some other
workingspace delivery from Vuescan could please your taste regarding
hues and saturation.

nikita

http://www.bergdesign.com/supercal/

 

[Bart van der Wolf]

SNIP

Unfortunately, I'm not sure. And I'm stuck on an LCD screen until I
can upgrade--soon, I hope--which introduces other limitations. Any
suggestions on how to do a passable eyeball test of the monitor's
accuracy without professional color profiling tools?

You can verify if at least the R+G+B gammas are aligned. Whether you can
adjust them individually depends on the display or video card controls.

http://www.aim-dtp.net/aim/calibration/index.htm Color Dither Gamma
Calibration Targets are very good. If you click on the "CRT Brightness and
Contrast..." link on the left, scroll down to the bottom of that page and
select the "Gamma-space evaluation is here" link, you will be presented with
targets for various display gammas. Something between 2.20 and 2.50 is
probably going to work with your display settings and will reveal if gamma
differences are screwing up your display of color.

If you really want to know your display gamma, I recommend
http://www.normankoren.com/makingfineprints1A.html . The target just a
little over halfway down the page will do a very good job of nailing it.

Bart

 

[Bart van der Wolf]

SNIP

http://www.bergdesign.com/supercal/

Mac OS 9 or X only!

Bart

 

[nikita]

I'm using a default manufacturer profile right now

It depends on in what way you're using it. How the conversion is made
and where.
And how that profile was intended to be used.

Good digital camera photos display
with colors that appear to be correct; if I can match their color
quality with the scanner I'll be happy. Is there any reason ICC
profiling the scanner would make things worse rather than better?

If the cam produces colors that looks good directly without any
manipulation it could mean two different things. The first is that the
monitor is in decent shape. Or that the cams colorballance just fits a
slightly screwed monitorballance. Compare with other monitors...

The intention with customprofiles is to get it better. However a
wrongly used customprofile – even if the best in world – is crap. It
depends very much on WHAT state was the scanner in when profiled. It's
a big difference in using profiles made within a internal workflow
like Vuescan /Silverfast compared to an external made profile with a
thirdparty solution. With the external solution, the scanner is
"optimized" regarding a gammavalue that sits best for some midparts of
the scannertarget. Then it is all locked at that setting. Then
profiled. All the real world scanns are scanned with this locked
setting. That is what the profile describes. That profile is ASSIGNED
in Photoshop and then CONVERTED to workingspace.

With Vuescan or Silverfast own creation of profiles, it's a floating
concept that is using a scannerprofile internally with all the
autofeatures on. You will then use the scanner as usual. The profiling
just ballances up a bit. The autofetaures and adjustments made by you
in the scannerstage is affecting the result of the finished mainscan.
With thirdparty profiles and conversions in photoshop, the same
delivery each time. If the orginal image that you put into the scanner
is light, the delivery is light. Dark, then delivery is dark. But the
scanner sees everything from black to white correctly. As you will
scan in 16bpc you can use photoshop as a scannersoftware for all the
adjustments to get it into visual taste. Using thirdparty profiling
externally means scanning as a mechanical process without any visual
adjustments in the scanningstage. Some of us feel that as freedom. But
it might just take an IT8 target for you to use in Vuescan to get what
you want.

Even if the monitor is uncalibrated, does it make a difference to the
initial scanner output? Does a user-created scanner ICC profile
depend on his monitor profile for accuracy, or does an inaccurate
monitor profile merely prevent him from making accurate postscan
adjustments to the original scan output? In other words, if the
original profiled scan output is viewed later on a profiled monitor,
will everything look right?

No, a monitor profile that is off spot, is not affecting the colors
from the scanner more than visually on the monitor (your monitor).
Each ICC profile describes each device behaviour. The CMS engine sits
in between those profiles and "translates colors from one device to
another. But it's very important that the workflow is set up
accurately. Otherwise source and target will be wrong. Yes, if the
monitor profile is false, then the preview will be wrong in the
scannersoftware. But again, it depends on how the CMS is set up. The
preview could be based on something else than a correct translation
(conversion) between scannerpace ( scannerICC custom or generic ) and
the monitorprofile. Also the scanner may deliver anything else from
the scanner into Photoshop workingspace than the correct workingspace.
Photoshop could be set up for anything else than the s-RGB that you're
saying the scanner delivers. Any missmatch depending on wrong set up
or poor scannerdrivers is affecting....

If a scannerprofile is in use somewhere in the flow, then check where
and how....
Either you let the scannersoftware make an on the fly conversion from
the scannerprofile to the workingspace within the scannerdriver. Then
deliver a known workingspace into Photoshop.
Or you scan without any profiles in the scannerdriver (generic) and
make the conversion in photoshop from the scannerprofile>workingspace.
But that takes a locked down scanner without any autofeatures
involved. The profiles you're talking about is probably attended to be
used in the scannerdriver in it's own internal colormanagment flow.

Let me put it this way;
If you set the scanner to deliver s-RGB or generic, it would probably
be the same as most flatbed today is dafult set to deliver s-RGB. If
photoshop is set to s-RGB it should not be far off when landing in
Photoshop. If so, take the file to another computer with a CRT monior
calibrated with Adobe gamma. If still screwed up, build a profile for
the scanner.

If using a Mac‚ download the Supercal and tune the LCD of yours.


nikita