Dear readers, dear LSI Support,
...this is a lenghty post, so please take some time to read.
Let me summarize my ideas regarding a successful and satisfactory
IT8 Calibration for slide scanning.
First a little excursion on the workflow from reality to the scanned
picture on your hard disk (input path). I will have to discuss the output
path (picture on hard disk to printed picture) as well, but mainly
to explain my view on the way how IT-8 targets are created.
Input:
------
1.) Reality, this is the "original" for most people and represents
the real reference that needs to be matched finally.
2.) The slide that took the information from reality and stored
it through the physical and chemical processes applied.
3.) The file that represents the picture in binary format.
Output:
-------
4.) The printer that puts the binary info available onto a
paper with ink or other processes or a slide with light.
5.) The monitor that shows the binary info available on screen.
6.) The slide projector and the canvas.
So now the interesting part is the way that information goes from
one step to the next one. Mind that in each of these steps specific
limitations of the process and/or the target format apply. The main
target of each process is to loose as little information as possible
and to have the result in a format that can keep most of the interesting
information. The latter may lead to results that are not technically
"ideal" but I not an expert in this (yet?).
OK, here we go:
a.) = 1.)->2.) This is simple thing we call "taking a photograph".
In times before digital photography you had to select the slide
material by subjective impression of how much you "liked" the result
after the development process. There was no automatic white balance or
other methods to adjust the process.
So it was extremely hard (if not impossible) to compare reality
with the result and you just compared your memory with the result.
Surely you noticed that different slide materials have different
representations of reality. I liked very much the FUJI RD100 due
to its vital colors and nice representation of the blue sky.
For the AGFA CT100 that I had to take occasionally I must say that
for my liking the brown was to dominant.
In short: All slides have a color cast!
b.) = 2.)->3.) This is the process we call "scanning". Here the
scanner takes a representation of the information found on the slide
and will put this information in an electronic format. Also the scanner
has items that influence the scan results like
- the used light source (RGB Balance)
- the sensitivity of the CCD (Signal to noise ratio, or "Density")
- the resolution
In short: Also the scanner may have a color cast
c.) = 3.)->4.) The "printing" is already at a rather late stage of the
whole process (and makes it difficult to print good pictures).
The data stored in a file need to be output on a medium that
will be viewed later. This needs to done in such a way that the
subjective impression of the viewer is similar to that of the
reality in 1.)
d.) 3.)->5.) The normal viewing process on your monitor. Here it is
important to note that the "white color" is set to a value close
to reality, mostly chosen is 6500K.
Also the monitor has deficiencies showing the RGB values exactly as
they are on file.
e.) 2.)->6.) The projection of the slide onto a canvas.
Here the light source (quartz bulb) and possibly the canvas
influence the viewing result. If you have a dimmable light-source
in your projector, the color of the light will vary dramatically
from red to orange to yellow. The usual maximum temperature of a
quartz bulb is around 3000K (so much lower than sunlight of 5600K).
Summarizing immanent "errors" of the transition processing and the
corrections we find the following:
a.)
error: slide-material
correction: none
b.)
error: scanner-individualities
correction: ICM input calibration (IT8 target + reference file)
c.)
error: printer-paper individualities
correction: ICM output calibration (with a calibrated scanner?)
d.)
error: monitor-individualities
correction: ICM monitor calibration (with a special measurement device)
e.)
error: quartz-bulb temperature color / canvas
correction: none (your eye makes the white balance adjustment)
A brief intermezzo on the "old" method of taking and viewing pictures:
If you look at a.) and e.) you may notice that most slides are
projected with a color temperature that is much lower than the reality
was taken on slide. So a color cast of the slide material in the
direction of higher temperatures (green or blue) may have been a
simple but intended "method" to compensate for his.
Now concluding to the IT8 creation and the use of the targets later
for calibrating a scanner:
Creating an IT8 target is simply printing an electronically created
IT8 layout to an output, e.g. a slide.
Since the output is not ideal as well, the resulting IT8 printout needs
to be measured and the values for each of the fields in the IT8 layout
need to be stored in a file so that later the values retrieved by a scan
process can be compared against the originally measured ones.
This is why the IT8 target needs an accompanying matching reference file!
Mind however that the scan-result of such an IT8 file will create a file
that represents the slide AS IT IS and NOT the ideal IT8 values!
Finally the following questions arise:
1.) Why is the IT8 creation not using a calibrated output device?
This (ideally) should make the use of a reference file unnecessary because
a calibrated output device should print an IT8 target where each of the
color fields has the same color value as the binary IT8 layout had.
2.) The reference file for an IT8 target could easily be used to create
a printer correction? If the reference file tells what the measured
values of the IT8 output fields are, the difference of each measured
value and the binary reference should give the printer correction.
3.) If the slide material has a characteristic color cast build-in, why
not supply correction-references for popular materials in a professional
software dealing with slide-scanning? This way the initial intended
color cast for slides can be eliminated in a comfortable way and the
transition into binary "white balanced" data would be very easy.
-> LSI please take that hint!
OK, I have my lessons learnt and could put together a complete and
logical model of the processes that I know of. This model explains all
the phenomena I came across. It is still simplified in that it deals
with color-cast only (and not much with non-linear response or other
peculiarities), hey but it?s just a hobby...
Thanks for reading so far and following my thoughts. I would really
welcome some feedback, from both the forum readers and the LSI
professionals.
bye
Tobias
, but I am curious 
