The Poor Man's ICC'ish Color Profiling

[Paul Verizzo]

OK, here are my iconoclastic thoughts about profiling. I'll probably put this up on a more permanent link.

PHILOSOPHY:

The poor man's ICC'ish color profiling.

I'm a graduate of the K.I.S. school of life's endeavors with a minor in Good Enough. In other words, how much time and effort and money is spent to tweak the last bits of whatever to perfection? (And then all that stuff like lighting variations, pigment ink problems, and human subjectivity undoes a bunch of that perfection stuff anyway.) Understanding the concept of diminishing returns, that with every additional input of time and money, the improvements are smaller and smaller. And that “stuff” is a poor substitute for skill.

Over the years I've collected quite a few step tablets, both physical and digital, or I scanned the physical ones. Besides any monitor adjustment you can do manually (and I have to say my Dell auto-adjust is pretty amazing,) a digital 21 step B&W tablet will instantly show if the highlights or shadows are being blocked. A well adjusted monitor will show a difference between each and every step. Got equality?

STARTING WITH MONOCHROME DENSITIES:

But you first need to make sure that the tablet is being seen as 0 in the black-most bit and 255 in the whitest. To do this, you need a photo/graphics program that will give the sRGB densities when you hover the cursor at a point on the step tablet. Canon Digital Photo Professional will do this, as will others. I presume Adobe products will, but I missed the Adobe gene, forgive me.

If you don't get 0 in the blackest rectangle and 255 in the whitest, you need to adjust the image by using black point and white point pickers in an appropriate program, and then save the image. The Kodak Color Portrait test print also has a handy 18% gray midpoint in a circle off of the woman’s face (and “perfect” Caucasian skin tone in another.) Use that portion for the "gray picker." You can download this image, with a monochrome B&W identical half that I combined at It includes a full range 0-255 adjusted 21 step tablet.

When you print that step tablet with your target paper and ink, you should see similar results, ideally. If not, you need different paper, ink, and/or profile changes. Canon printers, and I presume all others, let you make some manual adjustments which you can then save as a profile. Talking just B&W here, if either end of the spectrum is blocked, try reducing the contrast. Granted, this isn't an ideal Levels type curve, but it might work. Generally, I’ve found that if you have the full density range in the jpeg image, LCD monitors will show the gradations. Monitors are the least important in the lot of devices and programs used to make images.

ON TO COLOR:

This technique relies on the fact that scanners are usually equal opportunity devices. Very good at giving accurate colors and levels across the visible range, with little or no prejudice. Be sure any auto enhancements are turned off. Make a test print of the Kodak girl, and once again check the sRGB levels in the gray circle or the “139" tablet step. Same density. In the world of perfection, the hovering RGB data will be 139, 139, 139.

Oh, it’s not? Imagine that! Let’s pretend the numbers are 139, 101, and 151. That means the Red is spot on, the Green is weak, and the Blue is too strong. Since the Red is good, I’d bump the green by reducing magenta, and add yellow to bring the blue back into line. I sure wouldn’t worry about any numerical differences under ten, but YOCDMMV. (Your OCD Mileage May Vary.)

(Red is counter to Cyan, Green is to Magenta, Blue is to Yellow. Since all printers are subtractive, that means when you want to boost a given color, you subtract its compliment. And vice versa.)

There are two more criteria that you may choose to work with. Um, customize. One is Contrast. Increasing contrast will undo your hard work to get even densities, but for a given print, it may be what you want! The other is Saturation, what Canon calls "Intensity." You can have an image that has perfect RGB balance, but lacks color. It's weak and washed out. Bump the saturation. Er, Intensity.

So, experiment, and when you are happy (or attain “Good Enough”,) save that profile! Don't forget, no art exhibit ever had two similar prints sides by side to compare them! The eye/brain interface can adjust for a lot of objective sins! Stare at the color side of the Kodak portrait test image and then look at the monochrome. You will see an opposite hue in the B&W for a second! The reddish skin tone will be seen as cyan/green!

You can also get a double sided multi-colored “gray” card like this: http://www.ebay.com/itm/Pro-Photo-1...330?pt=LH_DefaultDomain_0&hash=item232a39f772 (If link is dead, just eBay search for “18% Gray Card,” and find the ones that are 5x7 inches with white, gray, black, and CMY on the front, and gray on the back.)

Do NOT use a scan of this or any other card and then figure “That’s black,” and “That’s white.” They ain’t, being reflective materials. They can never attain 0 or 255.

 

[RogerB]

OK, here are my iconoclastic thoughts about profiling. I'll probably put this up on a more permanent link.

Iconoclastic? From where I stand it looks like the opposite. You seem to be defending the Old Gods of Trial-and-Error, Skill' and Experience against the heretical ICC and CIE who preach that accurate colour can be achieved by control of the process using device profiles. I think the real iconoclasts are the ICC and I suggest that you are better described as a reactionary.

PHILOSOPHY:

The poor man's ICC'ish color profiling.

I'm a graduate of the K.I.S. school of life's endeavors with a minor in Good Enough. In other words, how much time and effort and money is spent to tweak the last bits of whatever to perfection? (And then all that stuff like lighting variations, pigment ink problems, and human subjectivity undoes a bunch of that perfection stuff anyway.) Understanding the concept of diminishing returns, that with every additional input of time and money, the improvements are smaller and smaller. And that “stuff” is a poor substitute for skill.

Over the years I've collected quite a few step tablets, both physical and digital, or I scanned the physical ones. Besides any monitor adjustment you can do manually (and I have to say my Dell auto-adjust is pretty amazing,) a digital 21 step B&W tablet will instantly show if the highlights or shadows are being blocked. A well adjusted monitor will show a difference between each and every step. Got equality?

STARTING WITH MONOCHROME DENSITIES:

But you first need to make sure that the tablet is being seen as 0 in the black-most bit and 255 in the whitest. To do this, you need a photo/graphics program that will give the sRGB densities when you hover the cursor at a point on the step tablet. Canon Digital Photo Professional will do this, as will others. I presume Adobe products will, but I missed the Adobe gene, forgive me.

If you don't get 0 in the blackest rectangle and 255 in the whitest, you need to adjust the image by using black point and white point pickers in an appropriate program, and then save the image. The Kodak Color Portrait test print also has a handy 18% gray midpoint in a circle off of the woman’s face (and “perfect” Caucasian skin tone in another.) Use that portion for the "gray picker." You can download this image, with a monochrome B&W identical half that I combined at It includes a full range 0-255 adjusted 21 step tablet.

When you print that step tablet with your target paper and ink, you should see similar results, ideally. If not, you need different paper, ink, and/or profile changes. Canon printers, and I presume all others, let you make some manual adjustments which you can then save as a profile. Talking just B&W here, if either end of the spectrum is blocked, try reducing the contrast. Granted, this isn't an ideal Levels type curve, but it might work. Generally, I’ve found that if you have the full density range in the jpeg image, LCD monitors will show the gradations.

If everyone throughout history had adopted your "near enough is good enough" approach we would still be living in the Dark Ages. A constant, conscious search for improvement is what sets us apart from other species.

Monitors are the least important in the lot of devices and programs used to make images.

Really?

Reading your various posts it is clear that you admire the great photographers of the last century. You will no doubt remember that Ansel Adams, one of the greatest, compared a photograph to a musical work. He said that the negative was the score and the print was the performance. In other words the actual making of the print in the darkroom was a large part of his creative endeavour. I suspect that had his prints been made in a high-street lab, they would hardly have attracted a second look. So, yes, skill, experience and vision all go into creating the darkroom print.

Taking the analogy to its logical conclusion we should say that Adams' print is actually a recording of the performance. The performance itself was carried out in private, in near darkness, and it resulted in a "master" record. To "play" the record as a public performance all that is needed is a suitable light source. The light source is analogous to the sound reproduction system for the (recorded) musical work. Its quality will influence the quality of the audience experience, but it does not alter the quality of the work.

If we apply this analogy to the digital realm then the image file direct from the camera is obviously the score; perhaps not so obviously, the performance is now the editing carried out in whatever image editing software we choose to use. Global or local adjustments of colour and contrast, conversion to B&W, cloning, sharpening, all are part of the creative process. All involve skill, experience and vision. The end result of the performance in the digital case is an image file, not a print - an image file that looks "right" on the photographer's monitor. The colour information is defined unambiguously, pixel by pixel, by the RGB values, within the context of the stated colour space. The image file is now the "master" record, but the difference of course is that we can make an unlimited number of perfect copies of this digital record.

It may be argued that the image can only be displayed as the photographer intended on the monitor used for editing; but, let's assume that the monitor reproduces colour accurately against a known standard. In that case any other monitor calibrated and profiled to that standard will display the image as intended. If this were not true then all photographic salons for projected images would be meaningless. How could the judges determine the quality of the submitted images?

The same argument applies to printed copies of the master record. Any printer that reproduces colour accurately, to a defined standard, can make a good print of the image. Out-of-gamut colours will differ from printer to printer but they are generally only as significant as small differences in bass or treble response in a sound reproduction system. They have little impact on the perceived quality of the work. Inkjet printing in itself is not an artistic activity!

ON TO COLOR:

This technique relies on the fact that scanners are usually equal opportunity devices. Very good at giving accurate colors and levels across the visible range, with little or no prejudice. Be sure any auto enhancements are turned off. Make a test print of the Kodak girl, and once again check the sRGB levels in the gray circle or the “139" tablet step. Same density. In the world of perfection, the hovering RGB data will be 139, 139, 139.

I hate to say this but the average scanner is not terribly good at reproducing colour accurately. Anyone who has used one to copy a piece of artwork will tell you that. But let's assume that it does give you accurate measurements. How do you print the test image? Do you let the printer manage colour or do you let the application manage colour. Either way you are using an ICC profile for the printer. So, to demonstrate that you don't really need ICC profiles you start by using an ICC profile? Why not eliminate the heresy of printer profiles and print with no colour management? I defy anyone to achieve even "good enough" colour from an inkjet printer using just the sliders in the printer driver.

Oh, it’s not? Imagine that! Let’s pretend the numbers are 139, 101, and 151. That means the Red is spot on, the Green is weak, and the Blue is too strong. Since the Red is good, I’d bump the green by reducing magenta, and add yellow to bring the blue back into line. I sure wouldn’t worry about any numerical differences under ten, but YOCDMMV. (Your OCD Mileage May Vary.)

(Red is counter to Cyan, Green is to Magenta, Blue is to Yellow. Since all printers are subtractive, that means when you want to boost a given color, you subtract its compliment. And vice versa.)

This is where your philosophy (method) goes a bit awry.Trying to correct RGB values by using CMY adjustments is an exercise in pure frustration. Your red is good but you want to bump up green by reducing magenta. Since Red = Magenta + Yellow you have just changed your red. Your blue is too strong so you add yellow. You've changed your red again. Oh, and since Green = Cyan + Yellow, you've changed your green too. This can go on for an awfully long time - printing, measuring, adjusting, printing - before you get a result that "looks" right. Even then, the real accuracy is unlikely to be very good.

Nevertheless, the real error here is in assuming that if you can get a single grey tone to be neutral you will also have achieved acceptable colours. Printer colour spaces are not that cooperative.

There are two more criteria that you may choose to work with. Um, customize. One is Contrast. Increasing contrast will undo your hard work to get even densities, but for a given print, it may be what you want! The other is Saturation, what Canon calls "Intensity." You can have an image that has perfect RGB balance, but lacks color. It's weak and washed out. Bump the saturation. Er, Intensity.

More printing, adjusting, measuring (eyeballing), printing.....

So, experiment, and when you are happy (or attain “Good Enough”,) save that profile! Don't forget, no art exhibit ever had two similar prints sides by side to compare them!

Actually all of the art fairs I have been to recently had (expensive) originals on the wall and more affordable prints (reproductions) in a browser. Anyone considering buying a limited edition print could instantly compare it to the original. Who would buy a limited edition print that didn't match the original?

Paul, you have described a "philosophy" that ignores monitor accuracy and involves an unspecified number of test prints and evaluations that may be nearly objective (scanner) or are purely subjective. All this to get a result that is "good enough". My iconoclastic view is this; calibrate and profile the monitor - maybe 30 minutes; print a profiling target - in my case two A4 sheets; measure the target and generate a custom profile - about 10-15 minutes depending on the profiling application I use. For a well-behaved monitor and printer the result is virtually guaranteed to be accurate colour reproduction. And yes, I can verify that subjectively and objectively. Call it OCD if you like. I call it do it once and do it right.

Apologies (and congratulations) to anyone that has got to the end. I didn't intend to make this post so long but I don't have the time to make it shorter. And I do have a bee in my bonnet about accurate colour.

 

[Paul Verizzo]

@Roger, I certainly don't have the time nor energy to make responses as you did to my posting.

I will say this about Mr. Adams: As you infer, if you remove his technical expertise, most of his images are pretty mundane. So, yes, it is OCD technical excellence that makes his images, well, OK, a lot of them, powerful. And it is my opinion, repeat, opinion, that a lot of getting everything "perfect" is often a substitute for making powerful art, hoping that the latter will be achieved by the former.

OTOH, photography is loaded with many respected and famous ones that sent their film to a lab. I think I could name a few, but since I'm not positive, I won't slander them inadvertently.

The power of those images is in the image, not the perfection of tones.

"Good enough" is indeed "good enough" for many endeavors, none of them having to do with evolution. No, I don't want "good enough" during my brain surgery. But if I'm painting my house prior to sale, "good enough" is quite adequate. Don't need that 20 year paint, perfect prep. Maybe I react to perfectionism because as I got older I recognized my father suffered from that........and that I did, too! So, for my own better mental health, I try to identify the places and times where "good enough," is, well, "good enough."

 

[mikling]

A lot of people misunderstand the real value of ICC profiles. At the core is that it allows the artist to manipulate an image to the point that a good representation of the final product is possible.....and that is softproofing. This is lost on many.

Too many people look at ICC profiles as the cure to a good print. Not so. It is kind of a byproduct.

Now here's a simple analogy. Ansel Adams used his own eyes to adjust his prints in the "room". Now let's suppose he wore sunglasses that had variable densities inconsistently in the lens of the sunglasses. He would would not know what the end product would look like till it was finished. Calibrating a monitor attempts to remove the sunglasses. So his eyes are a direct path to the output.

What Paul is attempting to do is to use a set of sunglasses, he calibrates this set of sunglasses by matching the output so that what he sees is what is printed. This is theoretically the same as the profiles. The only problem is that his "sunglasses" is not referenced. Suppose he was to lose those sunglasses. Now what? ICC standards have set baselines. Everyone calibrates to that reference. So sunglasses are standardized.

Now there is a reference. Sunglasses ( monitors) attempt to replicate that standard.

The problem with Paul's method is that the limited amount of adjustments that can be made.

Now this takes me back to my tech school days where I had a lab exercise that was teaching us about engines and carburetors. It instantly became clear to me that mapping engine and fuel characteristics was really more than two dimensions. A simple carburetor and distributor could not do the trick to optimize an engine. I mentioned this in the late 70s to the lab instructor and I was given a strange look...like I was crazy. Fast forward and Honda VTEC was born over a decade later........I digress. The simple controls in the monitor and printer is unable to compensate for the non linear behavior in some circuits that a digital adjustment is capable of. If you look at some of the compensation curves that monitors need, you will come to realize that fine tuning is needed. This is best done digitally either by a direct mapping function or using some math functions that refer back to a matrix.

Take for example Argyll CMS when it is calibrating a monitor. The amount of iterations and curve fitting it is doing is unseen. You cannot easily get a circuit to replicate this curve. Yes Curve fitting math is quite compute intensive and the amount of data crunching it is doing to do a best fit is amazing. It is doing things mathematically, that I could not dream of in the 70s and would have thought not possible because of the amount of crunching that would have been required. For someone who has used Argyll, can you imagine the amount of time it would take on the original IBM PC or Apple II?

Long story. But profiles is really to see the end output...not to get perfect output. This is often misunderstood. You can have crappy gamut and rough volumes but the image matches the intention of the creator and that is what counts in color management. So it matters not whether you have an uber monitor and a 24 color printer or simply an orange monochrome monitor ( Amdek remember those?) and a laser printer. If you can "see" what the laser will print reliably, then it is color managed!

 

[Smile]

I agree with the @RogerB and with @mikling you wrote what I was going to write. I might add that given paper costs the cheaper way is to have ICC profiles created by buying calibrator or using such service then experimenting with unknown result. If it were the opposite nobody would use ICC profiles.

If one uses OEM ink and expensive paper like Canon Luster or Platinum, the ICC profiles pay for themselves with the amount of ink and paper they save otherwise wasted in testing and reprints.

Since any 4ink printer can print better in color then photolobs like fuji, kodak etc. one must be crazy to use such services. The photolabs however can print perfect BW without any profiles (by internal RGB calibration) so it's again crazy to use all black ink setups to get perfect BW prints.