Scanner-Kalibrierung mit X-Rite

[markmaus]

Hallo zusammen,

ich nutze folgende Kombination:

Epson V850 Pro
Silverfast 9 SE Plus
X-Rite i1|Scanner

Ich wollte mit x-rite gerne ein Spezifisches Profil erstellen. Leider habe ich dabei so meine Probleme.

Es heißt, man müsse die IT8-Targets ohne Bildoptimierungen und ICM scannen und als unkomprimiertes TIF abspeichern.
Das habe ich mit verschieden Programmen (Epson Scan V2, Silferfast SE Plus 9) probiert.
Bei Silverfast kann ich HDR RAW auswählen und dann speicher Silverfast die Bilddatei ohne Optimierungen und mit einem Gamma von 1 ab. Die Datei ist also ziemlich Dunkel. X-Rite kann damit nichts anfangen und bemängelt, das Gamma sei zu hoch.

Bei Epson Scan kann ich ICM auch ganz ausschalten, erhalte dann aber ein ebenso dunkles Bild.
Dann kann ich dort "Kalibrierung Epson" einstellen und dort ist dann wieder ein Bildschirmgamma eingestellt. Dann werden die Scanns heller und von x-rite akzeptiert. Die richtigen Referenzdateien zu den Targets liegen mir vor und ich wähle diese auch aus.
Allerdings wird immer bemängelt, das es Probleme mit dem Target gäbe und man darauf achten solle, das auch der Graustufenverlauf sichtbar und unten ist. Das ist jedoch der Fall.
Die Profile sind auch nicht brauchbar. Scanne ich z.B. in Siverfast ein Bild und wende das Profil an, sind die Farben total daneben, gehen z.B. ins grünliche. Wähle ich das passende Standardprofil in Silverfast 9 (z.B. "Epson Perfection V800/850 - Film") sind die Farben so, wie ich sie erwarten würde.

Was mache ich falsch, bzw. wie mache ich es richtig?

Danke für Eure Hilfe

Mark

[Joe]

Vorab: ich habe von der Software keine Ahnung.

Aber der Betreiber dieses Forums bietet eine Schritt für Schritt Anleitung für X-Rite an. Die findest Du hier: https://www.filmscanner.info/Scannerkal ... #i1Scanner

[Jossie]

Hallo Mark,

Du scannst Aufsichtvorlagen, vermute ich mal.

Epson-Scan kenne ich nicht, dagegen kann ich Dir vielleicht für SilverFast helfen.

SilverFast benötigt ICC-Profile, die aus dem gamma-korrigierten Scans erstellt wurden. Dir HDR-Formate sind dafür nicht geeignet. Du speicherst also ohne jedwede Bildbearbeitung und Farbmanagement ausgeschaltet in TIF. Wie ich bei meinem x-rite sah, soll man IT8-DIas mit 800 ppi scannen, damit die Schnittmarken automatisch gefunden werden. Bei den IT8-AUfsichttargets gibt Dir i1 sicher eine ähnliche Empfehlung.

Frage: Warum verwendest Du x-rite i1? Sehr gute Profile kannst DU auch mit der freien Software Argyll erstellen. Sie sind viel besser als das, was SilverFast liefert.

Hermann-Josef

[Joe]

Silverfast SE und X-Rite i1 sind (soweit ich weiß) im Lieferumfang des Scanners

[Archivist_Goals]

"SilverFast requires ICC profiles created from the gamma-corrected scans. HDR formats are not suitable for this. So you save in TIF without any image editing or color management turned off."

Hermann Josef, hey again. I'm circling back from earlier this month. Been talking with Franz Herbert of basiCColor for Input 6.

And he, along with my help, also tried to understand:

1. If SilverFast 12641-2 targets are up to the quality they claim when profiling with Input 6 instead.
2. If SilverFast IT8 Auto Calibration is reporting real dE values.

Historically, whenever auto profiling, I always obtained an average value at 0.9 and found this odd. I say it is odd, because others over the years have questioned the validity of the advanced chart layout, and how exactly LSI decided to calculate this in SF. (We talked about this a few weeks back in my original post.)

For example, I supplied Franz with a RAW scan from SilverFast that was outputted to 48-Bit HDR RAW format, TIFF, 1200DPI. Trying to import into Input, along with the target's data file, reveals a dE value that is way too high, ~20. This only works on gamma 1.0 images that have been gamma-corrected. But then, when you correct the gamma, they are no longer 'raw'. The same results occur with running this test with VueScan, outputting to a 48-Bit RAW-format file (not 'TIF', but VS' 'RAW' format.)

So, none of this makes sense. We're supposed to be profiling a scanner with a profiling software and using a linear, RAW-AS-ONE-CAN-GET, scan to do this with. Any corrections to the RAW file, changes it not to be 'raw' anymore.

When you say that SF requires ICC profiles to be created from a gamma-corrected scan. Well, this does not make sense, since one is supposed to use a completely RAW (linear) scan to profile with from the outset. And since it would be RAW, it must contain a gamma value of 1.0 and thus would be very dark.

"So you save in TIF without any image editing or color management turned off"

- Did you mean to say, "Without any image editing or color management enabled?"

I've just about given up with using SF's Auto IT8 Calibration feature, and using their advanced targets. I am planning on investing in an HutchColor (HCT) target and purchasing Input 6 Pro. Input is used in serious museum-quality workflows. And the lack of transparency within SF, LSI, and how they arrive at such low delta E values is too opaque.

Thanks,
Nick

[Jossie]

Good evening Nick;

When you say that SF requires ICC profiles to be created from a gamma-corrected scan. Well, this does not make sense

Sure it does, depending on what the software using the ICC-profile assumes - gamma corrected or not. Gamma-correction does nothing to the quality of the data.

I've just about given up with using SF's Auto IT8 Calibration feature, and using their advanced targets.

As I said previously: Try argyll to create your profiles. The software is very good and it is free. It also supports the 12641-2 targets. Before you invest a large amount of money into other targets and an expensive software, you should try argyll and check the results carefully. You should also take a closer look at the features of colour management and ICC-profiles.

Here is what I get for my scanner and the extended target from LSI with argyll:

Profile check complete, peak err = 5.038859, avg err = 0.544388

Hermann-Josef

[Archivist_Goals]

"Gamma-correction does nothing to the quality of the data."

I would like to take your word on this, but I can't seem to find literature or other reference to this point online.

Let's say I create a scan in VueScan and output to 48-Bit RAW format. The scan will be linear. Creating archival scans implies that RAW is somewhere in the workflow. RAW, here, will not RAW like cameras produce. But it is 'raw' in the sense that it is capturing all possible information from the scanner's RGB filters. For archival purposes, we want all of that information to be captured and preserved.

If you alter the linearity of the RAW file, can you explain how this does not change the quality of the data, and can you point to information where I can cite this as fact? Not doubting you. But when I spoke with Franz earlier this week, he expected a completely 'RAW' scan from either SilverFast or VueScan, when imported into Input, to be accurately profiled without error (even though it has a gamma value of 1.0)

I should mention, that he did get it to profile correctly when the gamma was converted from 1.0 --> 2.2. But could not profile accurately (with a good dE) on 1.0.

Am I missing something here?

Argyll: Yes, I haven't forgotten about Argyll. I will test this out. But I am set on making some investments in better implements.

"Profile check complete, peak err = 5.038859, avg err = 0.544388"

This is a good result. But I want to switch to an HCT target, regardless. I've had too many people, including Wolf Faust, indicate that the 12641-2 layout's colors (some) are not possible to produce. I spoke with Hugo Rodriguez of EGM and he also concurs with this.

Thank you.

[Jossie]

I would like to take your word on this

One would not archive the gamma-corrected images if one is aiming for highest quality (see below). But this is not the issue here if we are talking about the optimum way to create an ICC-profile. It is only a matter of definition if the gamma-correction is incorporated into the ICC profile or if the profile is meant for gamma-corrected images. One should not confuse these issues here.

If you derive an ICC profile from a linear "RAW" scan, then the gamma-correction is automatically incorporated into the profile. You can readily use the "RAW" image in Photoshop and apply this profile to get a correct representation on you monitor. If you derive the profile from a gamma-corrected scan then you will get a correct image for display on your monitor with SilverFast. This is the difference.

There is a caveat with the gamma-correction: It is not strictly invertible, if we are talking about 16bit unsigned integer representation. Different values of the input image lead to the same value in the output image for the brightest pixel values. Gaps are produced for the lowest pixel values. You may verify this easily in a spreadsheet calculating the gamma-correction for low and for high values. But the gamma-correction is needed to produce images which can be viewed on a monitor...

Criticism from other manufacturers of targets are to be expected. My point is that the new targets do not cover a larger volume in Lab-space than the standard ones. They only provide a closer spacing of the measurement points. Thus the uncertain extrapolations in colour space are not removed.

Hermann-Josef

[Archivist_Goals]

Thank you for this explanation.

Right - for archiving, one archives the unedited, non-corrected, completely unaltered scan. Apologies if it seemed I implied the opposite. Because my goal is to create archive images. But from those archived-masters, derive the following.

1. A proper ICC profile that includes all of the information from the master scan of the color target derived from the scanner.
2. Map the ICC profile onto a RAW / linear scan of a scanned image and then further adjust in Photoshop, or other, etc.

The critics from others is to be expected, I agree. But where I am still confused is with the following:

Let's forget about SilverFast for a moment. About 20 minutes ago, I created a RAW scan using VueScan and one of Wolf Faust's regular reflective targets. The output from VueScan is a 48-Bit (16 bit) RAW file format (not to be confused with the 'TIFF file' output selection in VueScan; I only have the 'RAW' file creation checked off.) It it is a TIFF file. But just want to make the distinction. And with device RGB as output.

Using the trial version of Input 6 Pro - this allows one to use all of the features without limitation for 14 days - I imported into Input, with the Wolf Faust template provided in the software, imported the associated reference file for this target and it results in an avg. of 2.58 with a peak of 11.17 and std. deviation of 2.10. These result from setting set to Art Repro / Archival, Profile Optimization at dE 2000, Quality set to Custom, White Point off, Gray Edit on, Color Edit on.

So profiling to the above standard and settings ends in a failed profiled attempt - and this is because the imported scan still contains the 1.0 gamma value.

I don't want to go on about Input because Franz kindly spent all week helping me with this. But he came to the conclusion that one must change the gamma value - correct it to 2.2 - before profiling at least when it comes to Input.

Based on, "If you derive an ICC profile from a linear "RAW" scan, then the gamma-correction is automatically incorporated into the profile."

So then I must ask you, considering these higher-end tools are used in more serious workflows for digitization, how is one supposed to have the ICC profile automatically include the gamma-correction when the profiling software does not do this, while using a linear scan to create the profile in the first place?

SilverFast, in fact, did not do this, when scanning a target in their 'raw' mode (HDR RAW). If I profile a color target with 48-Bit HDR RAW mode - linear - the resulting profile created in SF cancels out the whole point of applying the profile to the raw image, because the ICC has the profile of a gamma 1.0 (which is why my scans always looked dark from a few months ago - I was going from 1.0 to 1.0, no change.)

I do not want to have to open Photoshop for example to manually adjust gamma. That would not be precise.

So, what step or software am I missing?

[Jossie]

So, what step or software am I missing?

I can only repeat what I wrote previously: Try argyll and there will be no problem in opening the scans in PhotoShop with the correct brightness. This is my experience.

Hermann-Josef