Calibration for printing

[adamhiram]

Thanks, @Tim Tucker 2, really nice explanation! In this case the paper used is Kodak Endura, which doesn't seem to have much of a cast to it as far as I can tell. I definitely missed the "Simulate paper & ink" checkbox, which wound up introducing a pretty even reduction in brightness of about 8-10% depending on where I sampled (255,255,255 went to 234,237,240), but no real cast. At least it's showing where the reduction in brightness might be coming from, and I'm glad you pointed out that extra checkbox for more accurate soft proofing.

I was doing some research and came across some interesting recommendations that may be worth sharing. In this video, the host comments that printed images almost always come out darker and uses curves to brighten them up slightly (just midtones?) before printing. In this one, a lab technician comments on trusting the lab to perform any necessary color corrections. Tony Northrup makes a similar recommendation in this video. It sounds like my best bet, if I don't want to spend a lot of time creating my own color correction export preset, is to simply check the color correction checkbox when ordering prints and let them do their thing. It's worked out pretty well in the past, I was just hoping to take more control over the process for more consistent results.

 

[Tim Tucker 2]

the host comments that printed images almost always come out darker and uses curves to brighten them up slightly (just midtones?) before printing.

LOL, it kinda misunderstands the whole system. Though colour is perceptual rather than absolute, (it's a bit of a chameleon in that is alters with it's surroundings), computers have to work in absolute colour. The RGB, LAB, HSL/HSB, or whatever colour space you use, co-ordinates describe a specific reference colour when viewed under the reference WB at light intensity *x*.

There is no point in calibrating a system if you're just going to *up the curves because the print is always darker*, which translates as *this file doesn't contain any absolute data as to the correct colour, I want you to print it like this so it looks the same as a 10% reduction in mid-tone luminosity and a similar perceptual reduction in saturation...* You see how the instruction quickly becomes ridiculous and meaningless, the only way you get your print to match the calibrated screen is to un-calibrate it, guess.

the "Simulate paper & ink" checkbox, which wound up introducing a pretty even reduction in brightness of about 8-10% depending on where I sampled (255,255,255 went to 234,237,240), but no real cast. At least it's showing where the reduction in brightness might be coming from, and I'm glad you pointed out that extra checkbox for more accurate soft proofing.

This can be misleading, and confuse many. To be clear, the colour co-ordinates of 255;255;255 represents pure white when viewed under a reference WB light source at intensity *x*. Your soft proof is *not* showing you how the computer converts the colours as the target colour is still white, it's showing you a *visual simulation* of how the print may look different. The print will look different as it will have a lower contrast, the whites will look different because they will reflect the colour cast of the paper colour rather than an accurate reflection of the reference WB light source at intensity *x*. The point is to get as close to the *reference colour of 255;255;255* as possible and not print it as 234;237;240, (255;255;255 will be just the paper colour with no ink, if you add ink to achieve it then something is wrong...).

A slight magenta cast works in this way; it's not a real colour in the sense that there is no wavelength that corresponds to magenta, but a combination of red and blue, (do not make any correlation between RGB co-ordinates and quantities of red, green and blue. Light is not RGB, they are simply co-odinates in a 3D colour space, nothing more). Adding blue to yellow in a subtractive space will darken it and de-saturate it, the red component will shift the colour towards orange, or if a very pale yellow will shift it towards pink.

You also have to be careful with computer screens as the additive colour system is most prone to being corrected or altered by the eye, colour casts are harder to spot on computer screens than they are in print. The only way to assess colour accurately is to compare the colour on the print to a real Pantone swatch of the closest reference, what you're really trying to do with calibration is to get your screen to do the same, or as close as possible.

Now caveat first: Do not rely on my accurate assessment here as I do not have prefect colour vision, I have a moderate form of the common colour-blindness. But...

If I refer to your original JPEG from flickr and take a close and careful look I see a slight warm cast to it. It's more visible in the shadows and highlights. If I offer a slight correction on my calibrated Mac I get this, which has more neutral shadows and highlights, your original on the left:

In defence of printers; any one worth their salt knows how deceptive colour is on a computer screen and are also aware that most of us suffer the delusion that our vision is absolute and simply don't consider that it's us who don't see it correctly. So the colour corrections they offer are normally geared towards removing casts and producing neutral skin tones. They guess at what we think we see when we present the image for print and correct the things we don't see because they know it will become far more visible in print. These are their references and not the colour of the shirts because, a) they have no reference of it, and b) most people don't notice details like that.

Unless of course you tell them not to...

Once you understand that vision is not absolute, (something I've always known), then you learn to look a lot more carefully and objectively. Then you see that things combine to produce errors that are largely unavoidable. For instance if you start with a slightly warm tone, print it on paper that has a natural slight magenta tint, (your soft proof suggests a slight cyan one but it is a representation of how the print will look once run through the calibrated process and not an accurate determination of the absolute colour which remains the absolute colour space co-ordinte contained in you file), then view it in a domestic house with warm tone lighting reflecting off magnolia walls... You see where I'm heading with this?

If you take a peek across at another popular forum you will see quite clearly how little some photographers understand colour, they simply don't see it or the mistakes they make. So have a discussion with your printer with an open mind because your prints may have a slight warm cast, their calibration might not be spot on in the same way that your's isn't. They might always remove slight casts even with *no correction* selected because they've looked at the other popular site as well. They will always try to give you what you want but you must always accept that there will be variations. As indicated their prime concern is judging by skin tone, your's by the shirt colour, that alone will produce differences.

I can't give you a solid answer because I'm just judging on a calibrated screen, I've not seen the actual print but only, if you like, a soft proof representation of the colour difference you see. There are no absolutes in it to measure. Hopefully I've given you a fair appraisal of the difficulties that will allow you to evaluate from a more fluid model rather than making absolute assumptions of the colours you see.

Addition: If I look at your original I see that it doesn't have a true white point, all the colours are slightly darker. Now this is no indication that your image is under-exposed, it looks to be well exposed. But consider it in terms of absolute colour rather than the perceptual brightness of your screen. Yellow is a perceptually bright colour, it lives in the high numbers of the RGB colour space. The lower numbers you interpret as bright pale yellow on your screen actually are slightly darker with a red tint, the target RGB co-ordinates that describe that actual colour to be printed will render from slightly more orange to pink depending on the saturation, (the whiter pale yellows tending to show more red). Your print will always have less contrast and in low light will not be as bright as your screen, especially if the screen brightness is high. White points are not always desirable in prints as long as you're close, but do be aware that bright pale yellows are not readily achievable without that white point, you will see a difference. It's the trouble with actual vs perceptual and computer screens. A screen will always look bright as it generates light, a print needs to reflect light so if you add a slight density to the paper base it darkens the colour and changes your perception of it, noticeable in the bright yellows and fairly invisible in the blues.

 

[adamhiram]

Hopefully I've given you a fair appraisal of the difficulties that will allow you to evaluate from a more fluid model rather than making absolute assumptions of the colours you see.

Thank you for the great explanation of these challenges, I can certainly appreciate that there are a lot of variables that can affect the perceived colors in a print. So what do you recommend for getting more accurate colors when using a printing service? As you can see here and here, when I used Mpix and used their color correction services, the colors and overall brightness came out pretty close to what I expected. Am I better off just checking the "use color correction" checkbox, or is there a better alternative?

 

[Tim Tucker 2]

Am I better off just checking the "use color correction" checkbox, or is there a better alternative?

Sometimes, yes. With the shots I've seen there is a very clear indication of what you require, the trouble with many shots edited on a computer screen is that those doing the editing are not aware of how the eye deceives. If I download your file again, (I don't keep it on my computer), and have a look and see:

From this it is very clear what you require, but look at the histogram and the colour sample taken from what appears to be a bright white. This is the actual colour that it is as defined by the RGB colour space value:

As you can see it looks a little different from how you perceived it on your screen, I bet you saw it as more white with maybe a trace of pale yellow.

Next if I do a simple auto WB, this doesn't give you a *correct* WB but instead aims for a balance of primaries because that is the way your eye tries to correct it and goes someway towards what you think the colours are:

What the correction tells me is that there is a very warm cast on the image. So I deduce the following; you're after a straight portrait as rendered with good contrast and slightly warm colour to give a warm and pleasing skin tone. This is what you see on your screen, but it is not the reality that's described in the absolute colour of the RGB colour space co-ordinates. The difference here is probably just how your eye is correcting the brightness and colour of the image when you view it on a computer screen. In a print it becomes obvious, and more so if it combines with a slightly warm paper base or calibration that's slightly off somewhere.

I bet in time you will see this far more clearly on your screen, and about that time you will also see where the *tick colour correction* images are not quite what you want. This is because your colour vision will improve, you learn to see it more accurately with practice, (you'd be surprised just how far out it can be when you glance with an assumption that vision is absolute). Around this time you will be sending images to the printer and be getting back almost exactly what you expect.

I hope this helps.