Improvised color chart - How reasonable is this?

[Wolfram Malukker]

So I came across this image this afternoon, with the explanation that they forgot to order a color chart in time and needed one *right now* for work.

I figure that this *might* work, depending on how the paint chips were printed. Has anyone used paint chips to build a color chart, and compared it to an actual color chart?

 

[Kino]

For what use?

If you are just going to print out an image, just use the macbeth chart you reference.

Neither will be good for critical color work, but for casual use.... who knows?

A real, 18% gray card will be a better critical color reference than any printed out image. You can balance out the RGB and have true neutral colors...

 

[koraks]

For what use?

That's the $1M question.

Has anyone used paint chips to build a color chart, and compared it to an actual color chart?

Nope, but I'm already pleased that this apparently isn't just going to be inkjet-printed on a bog-standard CMYK printer. Right?
The main concern would be metamerism, so the question @Kino asks above is quite pertinent.

Looking at the patches - I see no saturated hues. It's all low-chroma. Is there any particular reason for this?

 

[DREW WILEY]

Buy a MacBeth Color checker chart and learn how to use it. It will be well worth the expense. Paint chips are better than nothing, but not even close to the real deal. And the samples shown in the diagram represent paint brands using low quality pigments incapable of hue purity, along with so-so printing consistency. The only advantage is that they're free.

MacBeth patches are very carefully made and critically balanced in terms of color purity and grayscale neutrality; paint chips aren't. Then you have the problem of metamerism, which Koraks just mentioned, which potentially plagues not only paint chips but paint itself.

 

[cmacd123]

I can understand the cost factor as I remember buying a set of 4 Kodak Grey Cards for under 10 bucks and even less for the small Kodak Colour Separation Guides with a reasonable Grey scale. Using actual chips from the Paint store - in the most saturated versions you can get might give you an internal reference, as long as they Chips will not fade during the time you are using them. the colours of course would not be calibrated to ANYTHING.

I have used an AGFA colour chip chart in the past, But that is no longer made as far as I know.

Looks like an 18% grey card can stillbe purchased as an Off brand for 20 bucks. more like 80 US for a single Kodak Grey card, which is proably what they would have cost back then if Kodak was not selling them in the hopes of selling more film and paper.

 

[Wolfram Malukker]

I have no idea why the colors were chosen, other than they were done by some computer somewhere. I saw this chart made and wondered just how accurate it could be, considering it was being used for a "I need it right now" color chart.

I mean, I would *like* to purchase a Kodak Grey Card Plus, but they don't seem to be available right now. The combo of the grey, white, black, plus the RGB color section seems ideal for what I would be using it for (learning how to edit and adjust my scans in Lightroom-I just got a Canon Canoscan 9950F)

 

[Kino]

This is Color Management and... I hope you know what you are getting into.

A chart will do you little good if your monitor is not correctly setup and profiled for the color space(s) in which you intend to print/publish.

The Internet is full of tutorials on this subject; you might want to read up a bit before deciding what you need to calibrate anything.

Here's a document from X-rite (of course, hawking their solutions), but it looks good for an introduction to the concepts and problems you will face:

https://www.xritephoto.com/documents/literature/en/L11-176_Guide_to_CM_en.pdf

If that doesn't deflect you, head over here and check out DisplayCal; a free color profiling software on par with most commercial applications, but not tied to one specific measuring device:

DisplayCAL—Display Calibration and Characterization powered by ArgyllCMS

Display calibration and profiling with a focus on accuracy and versatility

displaycal.net

I use this software along with a X-rite SpectraSensor Pro to generate LUTs for Rec. 709 color space on DaVinci Color Correction system.

The documentation alone is worth reading.

Good luck.

 

[Wolfram Malukker]

A previous employer had multiple color management systems for different tasks, an integration sphere, and an extremely expensive luxmeter in the lab. They were not my responsibility to manage, and I was not into photography at the time, or I'd have retained some of those items when the lab was closed and liquidated.

I am familiar with process management, and the work it entails-it's part of the curriculum I teach. Color management is just another process.

Thanks for the links. I'll check them out.

 

[Taylor Nankervis]

Paint chips!??


Though not cheap, a Macbeth colour checker (small or large) will pay big dividends in its accuracy of reference colour. I know they are very expensive (I have two) but stored and used carefully, they are better by far than any improvisation. Very especially paint chips!

[ This subject can open a can of veritable worms all-sorts, leading to things like monitor, printer calibration and profiles, media profiling, etc., etc. It isn't mastered in a day!! ]

What are you going to be using your (improvised) chart or a professional grade chart for?

 

[koraks]

(learning how to edit and adjust my scans in Lightroom-I just got a Canon Canoscan 9950F)

Oh, OK. Well, it'll be better than nothing, and it's a fun experiment, so go ahead; see what happens.

The combo of the grey, white, black

See how these are in real life, and under different lighting conditions. If you take 6 chips of paint ranging from white to black through grey, I bet you'll end up with 6 distinct hues, none of which are dead-neutral. Even commercially-made grey cards aren't perfect in this sense much of the time.

A more sensible approach IMO would be to get an RF-type IT-8 target here: http://www.targets.coloraid.de/ You can use that to build an ICC profile for your scanner. It'll work reasonably well - and overall will be better than your paint-chip approach because you can remove your own subjective assessment from the imaging chain (not to mention your computer monitor!)

 

[reddesert]

Looks like an 18% grey card can stillbe purchased as an Off brand for 20 bucks. more like 80 US for a single Kodak Grey card, which is proably what they would have cost back then if Kodak was not selling them in the hopes of selling more film and paper.

Slightly off topic, if one just needs an 18% grey card for amateur use, and doesn't mind a little age, there is one in many editions of the Kodak Darkroom Dataguide from the 1960s-1970s, along with a zillion other pieces of useful slightly dated information. The guides with black or brown covers are likely to have one (why do I have 4 or 5 Darkroom Dataguides? don't ask)

 

[Mr Bill]

I am familiar with process management, and the work it entails-it's part of the curriculum I teach. Color management is just another process.

Cool! Fwiw "Color Management," of the ICC style, is a bit of an indirect sort of thing. It's fundamentally based on measures of human perception, and various machines - digital cameras, etc. - do not necessarily match same. It's kinda overly complicated to explain here.

Let me give a very brief (for me) explanation... in the late 1920s (I think) there were some "color-matching experiments" documented (look up Wright and Guild? ... going from memory). Essentially, they used three colored lights, adjustable in strength, and had various people to try to match a bunch of spectral colors by dialing up the strengths of their 3 colored lights. More complicated than what I'm saying. The results were summarized into a set of "color matching functions," where each spectral color (defined by wavelength of its light) has three values, one for each of the colored lights. Some of the values had to be negative, roughly meaning that the color could NOT be matched, except by putting an opposing color on top of the spectral color. Now, it turns out to be mathematically possible to recalculate the color matching functions for different colored lights, and even for hypothetical, but impossible, colored lights. So, apparently in order to eliminate negative values in the color-matching functions, they invented some "impossible" colored lights, so-called "imaginary primaries." So today's commonly used color-matching functions are based on same.

With that background, let's say that you want to determine the "color" of some given test patch, whether it's Macbeth ColorChecker or your own paint chip. What you would ideally do is to first measure the spectral reflectance of the test patch through a range of roughly 400 to 700 nanometers, the main sensitivity range of the human eye. (This is done using a spectrophotometer.) You also need to know the spectral makeup of the light source used for viewing the test patch. Finally, for each wavelength you would multiply the light source times the spectral reflectance, and then times the CIE color-matching function. This gives a set of CIE coordinates which, after normalizing vs a perfect white reflector, can then be converted into something like CIELAB values (L*, a*, and b*). CIELAB uses, as I recall, a so-called D50 illuminant, a hypothetical source roughly equivalent to daylight with a color temperature of 5000K. And yeah, I know how to (or at least used to) do the calculations.

Now, I note that your sample image of a Macbeth ColorChecker has a set of three numbers, presumably some version of RGB in 8-bit, 0 to 255, numbers. Now, those numbers are actually meaningless without specifying a so-called "color space" for the RGB system (typically this space might be what they call sRGB, where the primaries are defined, as well as an illuminant (I think D65 for sRGB) and a tonal response from nominal black to "white").

Having put up my set of smoke and mirrors, I would now ask, do you think you know what the values of the Macbeth ColorChecker are? So this is the sort of thing you'd be dealing with when using a ColorChecker to assist in getting the colors "right." It's a fairly big learning curve.

I'll make another post regarding the actual (original) Macbeth ColorChecker.

Also, glad to answer, or at least try, any questions, within reason.

 

[Mr Bill]

When the Macbeth ColorChecker first came out, around 1976?, I was a young wet-behind-the-ears former shooter (photographer) moving into lab work to learn about color processing. As a QC tech one of my information sources (later a good friend) was a fairly recent RIT photo science grad, in a related department. He and my department manager made the case to equip the company studios with Macbeth ColorCheckers. Enough of them that Macbeth couldn't supply immediately.

At the time there was a technical paper published, describing the design of the ColorChecker. It was titled something like "A Color Rendition Chart...." or something like that, with C.S McCamy as a principal author. It's probably still available on the internet. (I have occasionally posted a link, but I kinda doubt that anyone here reads it, so I seldom bother anymore.)

Some main points of the Macbeth ColorChecker were to attempt spectral matches, as best as possible, of some common things, including a couple of flesh tones. (The tech paper showed a number of curves comparing the real thing vs their results.) The idea being that a film tester could photograph a typical foliage, or whatever, out of season, and have a common reference with other photographers across the world. (In those days, fwiw, the reproduction of color was nowhere near as good as it later became, so such things were pretty important.)

So if one tries to make their own color test target, it's extremely unlikely that they will be able to match the spectral makeup of the Macbeth ColorChecker. So for example, if one changes the light source it's unlikely that the two charts will shift in the same way. And visually, they might match under one light source, but not under another. This is called a "metameric failure." A good example of this is in those color patches used to check for a D50 light source - there are two color patches that visually match under D50 light. But not under other lights.

Anyway, a self-made color chart could function as one's own in-house reference, keeping in mind that a light source change might have some unpredictable changes, etc., etc. And they might photograph differently than they appear visually under different light sources. The real Macbeth ColorChecker is gonna be mostly well-behaved under various lighting conditions.

So.... just depends on what your intentions are.

 

[koraks]

At the time there was a technical paper published, describing the design of the ColorChecker. It was titled something like "A Color Rendition Chart...." or something like that, with C.S McCamy as a principal author. It's probably still available on the internet.

It's apparently still hosted here: https://home.cis.rit.edu/~cnspci/references/mccamy1976.pdf
Thanks for highlighting it; looks like an interesting read!

 

[Romanko]

I figure that this *might* work, depending on how the paint chips were printed. Has anyone used paint chips to build a color chart, and compared it to an actual color chart?

I know people who make their own color charts but they are world-class printers and color consultants with PhD degrees.

A proper color chart has three important properties: (1) the color values of the patches are known and accurate; (2) the reflection spectra are known and the spectral curves are nice and smooth; (3) the charts are stable.

Nothing stops you from picking (often free) paint swatches from a hardware/paint store, gluing them to a piece of black cardboard and measuring the swatches with a spectrophotometer. I am not sure how stable the swatches are but you can always find it out.

If you tell us what you intend to use the color chart for we might be able to provide more help.

 

[Mr Bill]

A proper color chart has three important properties: (1) the color values of the patches are known and accurate; (2) the reflection spectra are known and the spectral curves are nice and smooth; (3) the charts are stable.

Hi, I would just point out potential issues with your #1, the "color values." In the normal ICC color management scheme, the known color values would be expressed in some sort of CIE terminology, meaning visually based, and only under the specified illuminant. Otherwise mostly agree.

Drew Wiley has occasionally pointed out potential issues even on measuring spectral reflectance. Where perhaps color changes when viewed at a slight angle, so even this is not foolproof.

The sort of stuff that I've worked with has been mostly fairly straightforward so never really dealt much with the tricky issues.

 

[Romanko]

In the normal ICC color management scheme, the known color values would be expressed in some sort of CIE terminology, meaning visually based, and only under the specified illuminant. Otherwise mostly agree.

D50, 2 degree viewing angle, XYZ and/or L*a*b* are commonly used "color values".

 

[Mr Bill]

D50, 2 degree viewing angle, XYZ and/or L*a*b* are commonly used "color values".

Yes, but as l sorta indicated earlier, in my long-winded explanation, but without specifically saying it, CIEXYZ and CIELAB are both based on human vision. That is, using the experimentally determined color-matching functions.

So a "limitation" of such a designation is that multiple spectral makeups can appear the same to a human, and have the same CIELAB values. Yet to a machine/digital camera might appear substantially different. So, as you sorta indicated earlier, a well-behaved spectral makeup would ideally be used in the color target.

 

[DREW WILEY]

Imagining that one can download a color chart of any serious value and print it themselves is a pipe dream. I don't care who they are. Understanding CIE color mapping is just the tip of the iceberg. There are all kinds of logistical hurdles after that. Where are you going to get the relevant balanced pigments? No inkjet machine can do it. No paint store machine can do it. If you want to try to reinvent the wheel, it's up to you. There are already corporations which have invested a great deal of time and money to provide accurate color aids. Might as well use them.

There are all kinds of features built into the MacBeth Chart you might not be aware of. The primaries (RGB) and secondaries (CMY) are very precisely balanced and equally saturated, so that any imbalance in reproduction (say, in a color film test or sample print) becomes readily evident. Then the tertiaries are well thought out, analogously. The gray scale is quite pure and neutral, with precise spacings, and with the middle patch providing a true 18% gray sample (something quite rare in so-called gray cards, which generally suffer from poor quality control). The sheen is close to ideal for various purposes. Just keep these clean and properly stored, and they last a long time.

Then, alas, there is the issue of appropriate lighting when viewing results, as well as when testing films.

 

[bernard_L]

A proper color chart has three important properties: (1) the color values of the patches are known and accurate; (2) the reflection spectra are known and the spectral curves are nice and smooth; (3) the charts are stable.

(my emphasis)
So you found on the interwebs a chart and intend to print it. Assuming your intent (not explicitly stated) is, for instance, to profile a digicam, or a scanner, or... the profile accuracy (or rather, errors) will just be those of your printer and its profile.
It is not essential to replicate the actual Macbeth colors; what is essential is to obtain --see post by @Mr Bill -- is the actual colorimetric values of each patch and generate a file such as:

NUMBER_OF_FIELDS 13
BEGIN_DATA_FORMAT
SAMPLE_ID XYZ_X XYZ_Y XYZ_Z LAB_L LAB_A LAB_B LAB_C LAB_H D_RED D_GREEN D_BLUE D_VIS
END_DATA_FORMAT
NUMBER_OF_SETS 288
BEGIN_DATA
A1 3.56 2.96 2.05 19.88 11.91 3.47 12.41 16.23 1.36 1.64 1.59 1.49
A2 4.36 2.95 1.70 19.83 23.71 6.96 24.71 16.36 1.22 1.81 1.68 1.45
A3 5.26 2.98 1.47 19.97 34.55 9.73 35.89 15.73 1.11 2.03 1.75 1.41
(....)

Affordable IT 8.7 (ISO 12641) Scanner Color Calibration Targets

Not me, just a satisfied customer.

 

[Kino]

The great thing about standards is there is so many of them...

 

[pentaxuser]

Well I looked at several Macbeths including the one at Cawdor. That's the Scottish one for those needing help and then decided to use the granddaughter's watercolour paints to replicate the Macbeth

I did a pretty good job if I say so myself. That should be good enough, don't you all think?

Replies on a postcard please. Any not delivered in the same earnest vein as my statement above will of course not be replied to

pentaxuser

 

[Sirius Glass]

This is a clever idea to use when on cannot get a Macbeth chart in time, but I personally will stick to the MacBeth chart for serious color work. Thank you for sharing.

 

[Wolfram Malukker]

Everyone here seems fixated on the idea that I would print this chart. Yes, the chart above uses the sRGB colorspace and hexadecimal color values, and at one point in my life I knew how to use them, but that's 25 years on now, and a consumer grade inkjet printer isn't going to print out that chart very well anyway.

Specifically, I said I am NOT printing a chart, but instead asked how paint chips compare to the "real" thing.

I would be using them to set up the combination of the consumer grade Epson inkjet, my monitors, and my scanner, to produce a consistent color chain. I'm not a specialized lab, I am not going to be producing anything spectacular, but I want less of a chance of being surprised when I order in a nice print from a professional printer. If I can scan in a test board with 8 paint chip colors, and they eyeball correct on the monitor, and they print reasonably closely on the printer, that's good enough.

I may also want to use them to play with spectral response in unusual films-but this is much easier done with a diffuse surface, a green laser pointer at 532nm, and a decent diffraction grating and some sunlight to produce an anchored spectrum.

 

[koraks]

I would be using them to set up the combination of the consumer grade Epson inkjet, my monitors, and my scanner, to produce a consistent color chain.

The sensible way to do this is get one of the IT8 targets @bernard_L and me linked you to, which will allow you to profile your scanner. This removes the uncertainty/variabiltiy of your monitor(s) and eyes from the chain. Then purchase any of the several monitor calibration pucks; they're not awfully expensive anymore. For the printer part, you can usually get very acceptable performance just using the OEM inks and a paper like Hahnemühle, and the appropriate factory-supplied profiles (this is one area Hahnemühle excels in, which is why I mention them). For cheaper papers, fall back on the ICC profiles the printer manufacturer supplies you with.

This approach will give you more than acceptable color fidelity throughout a scan/digital edit/inkjet print chain at very manageable cost. Moreover, the processes involved are repeatable and relatively simple, and you'll save lots of time that you can put into actual image-making.

If I can scan in a test board with 8 paint chip colors, and they eyeball correct on the monitor, and they print reasonably closely on the printer, that's good enough.

It sounds so simple, but the result will be disappointing. Profiling by eye just doesn't work well. You get some hues sort of right, and then when printing the next image you realize that some other hues have gone so far off the rails that you might as well have not even begun the whole effort. It's going to be a mess. Ask me how I know.

I may also want to use them to play with spectral response in unusual films-but this is much easier done with a diffuse surface, a green laser pointer at 532nm, and a decent diffraction grating and some sunlight to produce an anchored spectrum.

Yes, or just a plain prism. I like your idea of using a laser pointer to benchmark/quasi-calibrate the spectrum! You could even use a red and a blue one, which will give you two anchor points so you can interpolate & extrapolate the frequencies across the entire rainbow. Red and blue laser pointers are readily available; purchase ones for which the peak wavelength is known. Red will generally be 620-625nm and blue around 470nm.