Kodak grey card scan - green colour cast?

brian_mk · Nov 8, 2021

Mr Bill said:
You mentioned not being able to find (or use) certain ICC profiles. In case you don't know, ICC profiles also have an "internal name" in addition to the filename. And in my experience software generally uses the internal name. So it's possible, when you can't find a particular profile, that you are looking for the wrong name.

Yeah - I figured that one out using the Windows Colour Management Tool:-

I think you are right about the issues I am seeing being caused by the spectral response of the material being scanned.
It explains why the IT8 target printed on RA4 paper results in different colours after scanning compared to other neutral test targets I have tried (including the Kodak gray card and inkjet prints).

As I mentioned earlier, this makes me think that attempting to calibrate a scanner using a single icc profile derived from a target printed on one particular type of material is probably a waste of time. Manufacturers of Colour Management solutions don't tell you this - they just try to sell you expensive colour targets and calibration software.

As an experiment, I intend to photograph the Kodak gray card along with the WF IT8 target using my Nikon D810 using overcast daylight as the light source.
I will import the image into Photoshop and use the dropper tool to compare the RGB values of the neutral areas.
It probably won't help with the scanner calibration issue but it might be interesting to see the results.

RalphLambrecht · Nov 8, 2021

brian_mk said:
Yes it could be down to metamerism.

https://en.wikipedia.org/wiki/Metamerism_(color)

If it is, then it seems to me that trying to create in input icc profile for the scanner is a waste of time.
You would need different profiles for every type of printed source material that you scan.

So in my case the scanner may be calibrated for scans made using Fujicolor Crystal Archive paper (the same as the IT8 target) but anything else could give completely different results.

I see a similar effect if I make a print from the scanned IT8 target on Epson Premium Glossy paper and re-scan the printed version.
To my eyes, when viewed in daylight, there is no obvious colour cast on the printed version when compared to the original.
However, when scanned, the printed version has an obvious green cast.

I guess this is why you should not try to use a flatbed scanner to create or check icc profiles for a printer.
Presumably a spectrophotometer can make measurements that more closely match what the human eye sees.

welcome to color management. I once went to every supplier on Photokina involved in color management trying to find an expert. nobody ever claimed to know or be one. It is black magic and nobody wants to admit to that fact. In short: color management doesn't work and it never did! adjust to taste and then lock it in.

brian_mk · Nov 8, 2021

Just out of interest here is the result of photographing the IT8 Target and a Kodak gray card using a Nikon D810 outside in overcast daylight.
It was very hard to avoid reflections from the glossy surface of the Fujicolor Crystal Archive paper used for the IT8 target. Why use glossy paper I wonder?
I had to position the target away from the direction of the sun (even though it was behind thick cloud) and position the camera at an angle to the plane of the target to avoid reflections.

I used RAW and imported into Photoshop using ACR with Auto WB.
I added some gaussian blur to make it easier to compare visually.
The inner rectangle is from the neutral border of the IT8 target.
The outer rectangle is the Kodak gray card.

The Color Sampler Tool indicates the gray card as being pretty much spot on neutral gray.
The IT8 target is slightly blue (even though to my eye it looks slightly magenta).
These results tie up with the measurements I made earlier using my Macbeth densitometer.

There is no evidence of any green cast.

brbo · Nov 8, 2021

brian_mk said:
There is no evidence of any green cast.

Well, the rest would get green tint if you'd have white balanced on the #1 reading which is what in effect profiling on IT8 target does.

(I also have a number of WF IT8 targets (reflective and transparent), but maybe there is a reason WF targets are so (relatively) cheap?)

nmp · Nov 8, 2021

brian_mk said:
There is no evidence of any green cast.

View attachment 290548

Makes sense that you will get the complementary color of green (magenta) in the photograph. The icc profile will compensate by adding green to make it neutral.

I wonder if one can change the reference file of IT8 with own spectro - measured values which would presumably then give more accurate icc profile than the one provided by WF. Of course, it's a lot of work manually measuring all of those squares.

:Niranjan.

brian_mk · Nov 8, 2021

brbo said:
Well, the rest would get green tint if you'd have white balanced on the #1 reading which is what in effect profiling on IT8 target does.

I have to disagree with you there. The opposite of blue is yellow, so you would expect a yellow cast. That's not what I observe.

brian_mk · Nov 8, 2021

nmp said:
I wonder if one can change the reference file of IT8 with own spectro - measured values which would presumably then give more accurate icc profile than the one provided by WF. Of course, it's a lot of work manually measuring all of those squares.

Quite possibly. Unfortunately a decent spectro is quite expensive.

brian_mk · Nov 8, 2021

RalphLambrecht said:
welcome to color management. I once went to every supplier on Photokina involved in color management trying to find an expert. nobody ever claimed to know or be one. It is black magic and nobody wants to admit to that fact. In short: color management doesn't work and it never did! adjust to taste and then lock it in.

I used to work for Crosfield (later Fujifilm Electronic Imaging / FFEI) as a software engineer a few years ago before I retired, so I have a little understanding of the subject.
I did some work on scanners - although not directly connected with colour management, so I am not an expert.
There were a couple of gurus there with PHDs who did seem to understand this stuff.

Mr Bill · Nov 8, 2021

brian_mk said:
As I mentioned earlier, this makes me think that attempting to calibrate a scanner using a single icc profile derived from a target printed on one particular type of material is probably a waste of time.

Well, maybe. I don't think the RA4 paper dye sets vary a lot, so it might be that your Crystal Archive profile might come pretty close to Kodak papers, for example. But inkjets tend to be substantially different, so may not work so well. In fact, they (RA4 vs inkjet) will generally change appearance, compared to each other, under different light sources. With inkjet generally suffering the worst.

If you want to try this out, get two prints, RA4 vs inkjet, that nearly match under daylight. Then view under a compact fluorescent lamp. In my experience, up until 6 or 8 years ago, is that the inkjet is gonna suffer more. Why? Long story, but essentially the inkjets are capable of more color saturation. Using only three dyes, cyan, magenta, and yellow, you have to make narrower spectral peaks to get stronger saturation. But... the narrower peaks also make these materials more sensitive to spectral variations in the light source, meaning that they will tend to shift color more under different light sources, of which "energy-efficient" fluorescent lamps are something of an extreme example.

nmp · Nov 8, 2021

brian_mk said:
Quite possibly. Unfortunately a decent spectro is quite expensive.

For some reason, I thought you had one. I have a ColorMunki Photo which relatively cheaper but I didn't try it.

:Niranjan.

wiltw · Nov 8, 2021

brian_mk said:
As an experiment, I intend to photograph the Kodak gray card along with the WF IT8 target using my Nikon D810 using overcast daylight as the light source.
I will import the image into Photoshop and use the dropper tool to compare the RGB values of the neutral areas.
It probably won't help with the scanner calibration issue but it might be interesting to see the results.

I hope you take into consideration the color temperature of your light source under such conditions, is cooler than it would be in midday sun.

Mr Bill · Nov 8, 2021

brian_mk said:
I used to work for Crosfield (later Fujifilm Electronic Imaging / FFEI) as a software engineer a few years ago before I retired, so I have a little understanding of the subject.
I did some work on scanners - although not directly connected with colour management, so I am not an expert.
There were a couple of gurus there with PHDs who did seem to understand this stuff.

If there was a guy there named Abhay Sharma there, he is the author of a very good book, "understanding Color Management," or something like that. One of the best books, in my view, when the standard recommended books are not good enough. I've only seen the first edition, fwiw.

I was getting pretty deeply involved with making ICC profiles at the time, so I don't really have a sense of how difficult it is to understand for someone new to these things. But a lot of little issues I struggled with were very clearly resolved by Sharma's book.

Mr Bill · Nov 8, 2021

brian_mk said:
It was very hard to avoid reflections from the glossy surface of the Fujicolor Crystal Archive paper used for the IT8 target. Why use glossy paper I wonder?

You can get cleaner readings with an instrument from a flat surface. They'll be using a sort of optical layout, or geometry, to avoid specular reflections.

brbo · Nov 8, 2021

brian_mk said:
I have to disagree with you there. The opposite of blue is yellow, so you would expect a yellow cast. That's not what I observe.

The patch looked magenta to me. Anyway, you have the file, you can WB on that patch and see what you get, yellow or green. Or something in between (look, it rhymes)

I have a few WF IT8 reflective targets, a flatbed (Epson 4990) and a drum scanner. And a few profile makers and Xrite 310TR colour densitometer. Let me know if you can think of some test that I can run for you to help you troubleshoot...

brian_mk · Nov 8, 2021

wiltw said:
I hope you take into consideration the color temperature of your light source under such conditions, is cooler than it would be in midday sun.

Sorry, I don't understand. Why should the color temp of the light source result in different colour casts from a supposedly neutral gray subject shot at the same time?
Please explain.

brian_mk · Nov 8, 2021

brbo said:
The patch looked magenta to me. Anyway, you have the file, you can WB on that patch and see what you get, yellow or green. Or something in between (look, it rhymes)

I have a few WF IT8 reflective targets, a flatbed (Epson 4990) and a drum scanner. And a few profile makers and Xrite 310TR colour densitometer. Let me know if you can think of some test that I can run for you to help you troubleshoot...

Yes - it looked magenta to me too. However, my eyes are not as good as they were and the Photoshop tool rarely lies.

My conclusion (for now) is that the effects I am seeing are in part down to the accuracy of the WF IT8 target + reference file + the use of glossy Fujicolor crystal archive paper but also affected by differences in the spectral responses of different scanned media / paper types as pointed out by Mr Bill.
Having said that, I'm still open to suggestions.

Mr Bill · Nov 8, 2021

brian_mk said:
I have to disagree with you there. The opposite of blue is yellow, so you would expect a yellow cast. That's not what I observe.

That's the conventional view in photography, but I think there are other "color wheels" where that is not necessarily the case.

In conventional color negative systems, optically printed, the relationship of yellow vs blue, red vs cyan, and green vs magenta, are sorta locked in by definition. These systems are made to have complementary relationships, and the colors are sort of defined by the effect they have on each other.

If you look at another color system, CIELAB, for example, works with two opposing pairs of colors. Yellow vs blue is one, and the other is red vs green.

Something that was an eye opener for me, probably 1990ish, was when I decided it was time to finally nail down EXACTLY what "pure red", and "green," and "blue" were. I mean in terms of exactly what the defined spectral range was for pure colors. It turns out that there is no such definition; they are all somewhat vague, roughly that if something induces the sensation of "red" in the human eye then it is red. So again, in conventional color negative systems, we have, by definition, a "red-sensitive" layer that produces a "cyan" dye. So these are complementary, or opposite, to each other. So the definition of "red" essentially becomes a light that can affect the red-sensitive layer of the film. And the definition of cyan becomes the dye that results from an exposure to red light. So this is what I mean about the standard photographic "opposite" colors being established by definition.

So the exact definition of colors can be a sort of slippery thing.

brian_mk · Nov 8, 2021

Mr Bill said:
Well, maybe. I don't think the RA4 paper dye sets vary a lot, so it might be that your Crystal Archive profile might come pretty close to Kodak papers, for example. But inkjets tend to be substantially different, so may not work so well. In fact, they (RA4 vs inkjet) will generally change appearance, compared to each other, under different light sources. With inkjet generally suffering the worst.

If you want to try this out, get two prints, RA4 vs inkjet, that nearly match under daylight. Then view under a compact fluorescent lamp. In my experience, up until 6 or 8 years ago, is that the inkjet is gonna suffer more. Why? Long story, but essentially the inkjets are capable of more color saturation. Using only three dyes, cyan, magenta, and yellow, you have to make narrower spectral peaks to get stronger saturation. But... the narrower peaks also make these materials more sensitive to spectral variations in the light source, meaning that they will tend to shift color more under different light sources, of which "energy-efficient" fluorescent lamps are something of an extreme example.

Are you confusing Compact Flourescent with Cold Cathode Flourescent?
My Epson injket does not have three dyes. It has six.

brian_mk · Nov 8, 2021

Mr Bill said:
That's the conventional view in photography, but I think there are other "color wheels" where that is not necessarily the case.

In conventional color negative systems, optically printed, the relationship of yellow vs blue, red vs cyan, and green vs magenta, are sorta locked in by definition. These systems are made to have complementary relationships, and the colors are sort of defined by the effect they have on each other.

If you look at another color system, CIELAB, for example, works with two opposing pairs of colors. Yellow vs blue is one, and the other is red vs green.

Something that was an eye opener for me, probably 1990ish, was when I decided it was time to finally nail down EXACTLY what "pure red", and "green," and "blue" were. I mean in terms of exactly what the defined spectral range was for pure colors. It turns out that there is no such definition; they are all somewhat vague, roughly that if something induces the sensation of "red" in the human eye then it is red. So again, in conventional color negative systems, we have, by definition, a "red-sensitive" layer that produces a "cyan" dye. So these are complementary, or opposite, to each other. So the definition of "red" essentially becomes a light that can affect the red-sensitive layer of the film. And the definition of cyan becomes the dye that results from an exposure to red light. So this is what I mean about the standard photographic "opposite" colors being established by definition.

Sorry you totally lost me there!

So the exact definition of colors can be a sort of slippery thing.

Mr Bill · Nov 8, 2021

brian_mk said:
Having said that, I'm still open to suggestions.

I would just reiterate what I said before. The target that you have, the Crystal Archive paper print, is probably useful to characterize your specific scanner for use with other Crystal Archive prints. And since the visual appearance of that print is gonna depend, to some extent, on the the light source used to view it, the data file has to somehow accommodate this (the reference light source). A typical light source used, including CIELAB, I think, is called "D50." This is defined, spectrally, by a group called the CIE, and it essentially nearly the same thing as daylight with a color temperature of 5000 K.

So, for example, I'm guessing that your scanner "sees" a neutral part of the print as an R, G, B set of numbers. I said this part is "neutral," but this only means neutral under a certain type of light, which is likely defined as a D50 source, similar to sunny daylight. This is a sort of semi-common standard. So perhaps the data file says that a certain color patch has a CIELAB value that is neutral. And just to make up numbers, let's say that your scanner saw RGB values of 70, 80, and 90. So the ICC profile might essentially say, IF (70, 80, 90) THEN (80, 80, 80). Likewise the scanner would need to have every possible combination of RGB values translated to something else. Now obviously there are not enough test patches for this, so the system has to interpolate between measured points.

If you have made such a profile, and it is now in uses, consider what might happen if you now scan something that truly IS neutral, such as the Kodak gray card. By truly neutral, I mean that it has the same reflectance, near 18%, to EVERY wavelength of light. So no matter what the color of the light, it will remain neutral. So say that your scanner natively "sees" this as 80, 80, 80, but using the previous ICC profile (the one you made with Crystal Archive paper), it might now translate the actual (correct) values of 80, 80, 80, to something like 90, 8o, 70. Which we know is wrong.

This is essentially why the Fuji Crystal Archive profile is only "known-good" for scanning other Crystal Archive prints.

The whole thing is tremendously complicated under the hood, which is why most people don't go there. Generally they just want to learn the rules of how to use the profiles, etc. When I first learned about this sort of thing I was involved in a ground-up digital camera project (back when these were $25,000 systems). I was the color guy, and ICC profiles were relatively new back then; there was not that much information available, except for profile specifications, etc. So I dug into the guts of the thing. Today there's not much need for such an understanding, all the software/systems have become pretty stable, and mainstream. All you need nowadays is to know how to avoid getting into trouble, which generally doesn't happen. I think perhaps you expected too much from that target.

jtk · Nov 8, 2021

Kodak's grey card isn't "supposedly neutral." Never was. It was created and used to evaluate exposure and metering, had little to do with color.

Mr Bill · Nov 8, 2021

brian_mk said:
Are you confusing Compact Flourescent with Cold Cathode Flourescent?

No, in fact I don't even know what a cold cathode lamp is.

The compact fluorescents are those little curly tube things that screw into a conventional light socket. The reason they're so bad, universally, is not because they're curly, but because they were made specifically to be very energy-efficient, for that purpose of replacing the conventional tungsten bulbs (the kind that burn your fingers). The main way to be energy efficient with lamps is to put out light only in the wavelengths that a human is most sensitive to. So this sort of energy-efficient fluorescent mainly puts out a few spectral spikes, not a complete spectrum.

This sort of light is OK for a lot of things, just not for critical color in photographic prints.

brian_mk said:
My Epson injket does not have three dyes. It has six

That's a good thing, color-wise, if it has more. For example, a dye that is halfway between cyan and magenta, and another between magenta and yellow. But if it's something like cyan plus a weak cyan, and magenta plus weak magenta, etc., these don't really help the color problems that much.

The ultimate situation would be if you had enough different colorants that you could essentially make a spectral match to the original object you have photographed. Then it would act just like the real object under different light sources. But... that would mean that the camera would also need to similar sensitivities, not just the red, green, and blue values that we typically use today.

Part of all the magic of color photography is that we can make all these different colors from just three dyes. They essentially push the human eye's color sensing functions in just the right way to make the photo SEEM to have the same color as an original object. But the system sort of falls down when the print-viewing lamp is not correct, or when you use a scanner that doesn't "see" the same way as the human eye.

Mr Bill · Nov 8, 2021

brian_mk said:
sorry you totally lost me there [post 46]

Sorry, I get a little out of control sometimes. I tend to explain things the way the "younger me," say 20 or 30 years ago, would want to hear it explained.

When you are speaking to someone in person you can keep up with what they are understanding, or not. Across a forum, hard to know unless you know something about their background.

My whole point is that the "rules of color," what we learn in our younger years, just aren't always "correct." I used to be confused about the so-called "artist's primaries," red, blue, and yellow. When in photography I "knew" they were red, green, and blue. (How could the idiot artists be using the wrong colors?) Earlier you said that yellow and blue are opposite, or something to that effect. So how can this work with the artist's primaries? Are two of the primaries opposite (yellow vs blue), and the third primary, red, is all alone with no opponent?

It turns out that these systems all work out in their own realm, and part of the thing is that the lack of strict definitions for exactly what a specific color is leaves a lot of leeway.

Ps, if anyone wants a clarification, or simplification of anything specific, I'm glad to try to answer.

The main problem is that the underlying things are not that simple. In order to get a deeper understanding you really have to look at everything spectrally, which really complicates it. It's not unlike the case of driving a car, which can be readily learned as a series of rules that you can refine as you get experience. How hard to push the gas pedal and brake, and how much and how hard to turn the steering wheel. But if you want to get beneath the hood, to understand what happens when you push the gas pedal, etc., it gets much more complicated. Most people don't need, or want to, know these things. But if something goes wrong, and you want to solve things for yourself, then it may be worth knowing.

brbo · Nov 8, 2021

My WF targets behave similarly (Kodak one is better than those on Fuji paper):

Scanned in Vuescan with Epson 4990, WB in vuescan selected off the middle grey patch of the X-Rite calibration tile. It's night here so I can only look at my IT8 targets under poor lighting yet I can still observe the difference in greys in Kodak vs Fuji IT8 targets, but the difference is not as pronounced as in scans.

wiltw · Nov 8, 2021

brian_mk said:
Sorry, I don't understand. Why should the color temp of the light source result in different colour casts from a supposedly neutral gray subject shot at the same time?
Please explain.

Tints like Green/Magenta are on the other axis system of the color chart...warm/cool is on the yellow/blue axis of the color chart. While shooting with wrong warmth would not affect Tint, it does alter your perception of 'neutral'

Same shot but rendered Cool, Neutral, and Warm via about 400K in either direction from Neutral...about what you would experience shooting in Overcast conditions vs. in Bright Sun conditions.

Kodak grey card scan - green colour cast?

Privacy & Transparency

Privacy & Transparency