« Where Can You Buy Custom Profiles? | Main | Because Ken Complained... ** »

Monday, 27 July 2015

Comments

Ha, made me laugh to see this.

Now, since there are so many guru's lurking around, a question perhaps someone can answer... Like many people I have a well used copy of Lightroom CC and an underused copy of Photoshop CC. Is there any advantage to using Photoshop to print in terms of print quality? Did they throw in some secret sauce to the print engine?


[ Maybe Dave Polaschek or Andrew would like to answer this one. Hint, hint. - Ctein]

Yes, to Andrew's assertions... I totally concur.

And to the often very understated relevance of soft proofing... While the casual printmaker may not need soft proofing simulation of output image appearance when working with glossy and or luster print media that can come respectfully close to the tonal range we see in our digital images presented on a calibrated (factory or customized calibrator puck) monitor, soft proofing is essential, IMHO, when working with papers of limited contrast range, i.e, essentially all matt fine art papers. What are the best image edits needed to coax the best printed image onto the paper when the paper has such a constrained contrast range? You need the soft proofing turned on in Photoshop or other image ediiing software to figure out how to proceed unlrdd you waste several sheets of paper in a trial and error iterative image editing approach. That's a given.

So, as Ctein suggested in this interesting post "printer manages color" can work quite well if you intend to live in the deprecated printing world of sRGB or aRGB image source file color space printed on glossy or luster media with good overall Dmax properties, but if you choose to venture outside that comfort zone to use many other wonderful third party print media choices, then custom ICC profiles and soft proofed workflows are still going to win out, even with today's very well behaved modern printers like the Epson P600/P800 models.

kind regards,
Mark
http://www.aardenburg-imaging.com

My question is how does this apply to black and white printing or does it? I've toyed with the idea of getting a good printer and would like to know more about the steps, equipment, etc. for making small to mid-size (13 x 19) black and white prints. Does color gamut equate to gray scale? Any thoughts on tackling this topic?

Lightroom and Photoshop should produce identical output. I haven't tested this for a few versions but can. It's pretty easy to do; output the same profile targets to each application, measure, produce a deltaE difference report of each sets of measurements (I use ColorThink Pro).

I'll try this test tomorrow with LR6 and Photoshop CC 2015. Need ideally 24 hours for targets to dry before measuring so don't expect a report until Wednesday.

But again, unless there's an OS bug or application bug (and that's happened!) it should make zero difference using LR or Photoshop all settings being equal.

Than you for this refresher brief on while we shouldn't toss our printer profiles just yet, Andrew.

Still, I do plan to experiment printing without a profile on some of my most familiar (and often-printed) images. The last time I did this was around 2004 on an old Epson 2000. Everything printed greenish.

I saw the first article claiming the death of printer profiles and i refrained from being a naysayer. There's more to color correction than precise ink management. The DigitalDog straightened that and all is good with the world.

"don't mind suboptimal output"
I don't really understand this at all. How do you define 'optimal output'? Isn't the final result an artistic decision, one that cannot be defined as optimal or suboptimal?

Andrew's reminder about AdobeRGB clipping is timely and I agree with all three of his conceptual lessons. But one statement did surprise me a little:

... anything but ProPhoto RGB is suboptimal for output to the devices we're talking about

I thought ProPhoto RGB was much wider than most inkjet printers could use. Wasn't LStar RGB (eciRGB_v2 profile) supposed to be intended to deal with this situation? Is that correct, and if so, why isn't it used more widely? (I haven't had time to watch the video yet, so apologies if this is covered in there.)

I wish I found this convincing, but if you didn't see a difference, I find myself siding with your ability to see over "three conceptual problems".

Please continue the inquiry.

"Isn't the final result an artistic decision, one that cannot be defined as optimal or suboptimal?"

If it is your artistic decision to take a lot of information and randomly throw a lot of it away when making a print, then "yes." But imho, letting the machine decide which part of your information it throws away is making no artistic decision at all.

In agreement with Andrew, but my reason for custom profiles is that I print directly from RAW, I see no need for an RGB conversion before printing. When I need Photoshop, I use a smart-object for my RAW and do al the adjustments in adjustment layers, so I stick to the RAW-file as it is and print that.

To close out this discussion, it would be very interesting to repeat this exercise, comparing Prophoto and AdobeRGB.

I personally use Prophoto and Softproof when printing to matte papers, but keep getting criticised by people who insist I should work only in AdobeRGB, so would love to see this option included in the discussion, for completeness.

I use an HP Z3200 printer, this has an on-board photospectrometer and produces its own profiles for each paper, which I use with Qimage. In this case is there still any advantage in using the application based color management?

The main reason I've tended to avoid ProPhotoRGB is down to it sometimes picking up (UV/IR?) colour contamination on c41 film scans in the shadows, especially those taken in intense daylight - RA4 optical prints and aRGB scans don't show up any of this. for what it's worth, the film was NPH400 & the scanner a Hasselblad X5.

Apart from this debate, what I'd really like is an inkjet with proper (pantone matchable?) spot colour capability & a much finer dot in colour, closer to that obtained in B&W via piezography & QTR...

In Photoshop printing, one has to add output sharpening manually. Many of us use Photokit Sharpener plug in, from the Pixel Genius team. [Andrew Rodney is one of that crew.] For Lightroom printing, the Photokit output sharpener has been, to my belief, purchased and put into the Lightroom program. Lightroom automates much of the output in printing compared to Photoshop, but I believe the basic color and print control in LL are the same as PS, just with a different interface. I learned how to print from PS, with Photokit output sharpener, and I continue to do so. Both as a bit of a stubborn curmudgeon and only occasional fine art printer, but also because I just have not found the time to become facile in LL. And, yes, I use Prophoto RGB in all of my editing and printing because it has the color space to express what the camera captured and because the bit depth is there to withstand editing. Those who do it in 8 bit sRGB get less than the image can deliver. It would be mean, but maybe accurate, to say they get the image they deserve since they could have the best with just a few clicks on the keyboard.

A lot of useful information here, thank you Ctein and DigitalDog!
Can anybody comment on this:
If I want to make my own custom profiles without spending a fortune on hardware, is it worth buying a Colormunki device and use ArgyllCMS? I read somewhere that there is a method to use hundreds of color patches to create good profiles.
Is it worth it or should I use a profiling service? (Has anybody a recommendation for Germany/Europe?)
Btw. printer will be a new P600.
Thanks again,
Mark

Regardless of letting the printer manage the colors or using a custom profile, for my understanding using ProPhoto RGB as working colorspace for all tuning in Photoshop or any other 16bit capable software is just reasonable: Of course the color space gets reduced further down the printing pipeline (at the latest when the ink droplets hit the paper), but clippings or jagged histograms will be much more unlikely in a huge colorspace than in a limited one.

One thing I've learned from all this is the limitations of testing, even when I think I've controlled everything well.

The image I chose to print using Printer Manages Color appeared 99% identical to the print that used Photoshop Manages Color and an Eric Chan profile.

But I may have chosen the wrong image to test. I find that bad color profiles lose shadow detail and add banding to shadow gradients in my work. So I chose an image that would emphasize those areas. But those characteristics are a small part of what a good profile should do.

I also decided to test only on my favorite paper, Museo Silver Rag, but other papers, even using Eric Chan's profiles, may have shown completely different issues. You can't generalize about color management by testing a single image and a single paper/profile.

Anyway, I've got excellent custom profiles for half a dozen of my favorite papers, so I have no reason to complain. I'll keep using these profiles until I have to move to a new printer.

>>I don't really understand this at all. How do you define 'optimal output'?

Suboptimal: not optimal.

It's easy to define suboptimal and you can test this yourself after viewing my video on the subject. Suboptimal is output which visibly appears to show issues or problems! The first image I posted of two possible print processes show clear suboptimal output of the Blue ball using a canned Epson Luster profile compared to one that's more optimal by far: http://digitaldog.net/files/EpsonVsCustomProfile.jpg

Here's another example of suboptimal output, using Printer Manages Color instead of a custom profile: http://digitaldog.net/files/ProfilevsCtein.jpg

Nothing subjective here, banding, cross over of color purity, lack of shape (flat rendering), all signs of suboptimal conversions and thus printed output.

Does Lightroom 6 and Photoshop CC 2015 print the same? I have some results of the differences (or lack thereof) between printing in Lightroom 6 and Photoshop CC 2015. Printed the targets last night so they've had 12+ hours to dry.

Cut to the chase: The two are very close, I'd like to see slightly better numbers as I've seen in the past but when the rubber hits the road, you can print using either and get the same color appearance.

The geeky details, proof of concept:

For color geeks or the curious, I printed 1728 color patches through each application. You cannot print without color management in either Lightroom or Photoshop so I ran the patches through an output profile. Point is, apples to apple output, only difference is printing from two applications. This was done under Mac OS X 10.10.4. Epson 3880, Epson Luster paper.

I then measured the two processes on an X-rite iSis. The resulting spectral data was placed in ColorThink Pro to produce what is called a deltaE report. It's geeky but what's important is the average dE of the 1728 patches. Any value less than a dE of 1 is invisible to us, lower is better. Keep in mind that the instrument measuring the tagets isn't perfect. In other words, I could measure the same target two or more times and NOT get a dE of 0.00, there's always some 'noise' in the data. But again, when you see anything below a dE of 1, it's a visual match. The max dE (the one worst patch) is also interesting to examine.

Here's the differences between LR6 and Photoshop CC using the dE report from ColorThink Pro:

--------------------------------------------------
dE Report
Number of Samples: 1728
Delta-E Formula dE2000
Overall - (1728 colors)
--------------------------------------------------
Average dE: 0.87
Max dE: 2.81
Min dE: 0.04
StdDev dE: 0.40

Best 90% - (1554 colors)
--------------------------------------------------
Average dE: 0.78
Max dE: 1.38
Min dE: 0.04
StdDev dE: 0.31

Worst 10% - (174 colors)
--------------------------------------------------
Average dE: 1.66
Max dE: 2.81
Min dE: 1.38
StdDev dE: 0.27

--------------------------------------------------

The targets are untagged and in Photoshop, I set the RGB working space for sRGB so it would 'assume' those are the scale of the numbers. In Lightroom, untagged documents are supposed to be assumed to be sRGB. I want to actually Assign a profile to the targets and print again. That said, if the assumed color space between the two were different, the dE differences should be much larger. I think the dE differences I see here might be due to either dither being added in conversions (that would add more noise to the numbers) or a slightly different print path between the two applications. But with an average dE from 1728 patches being 0.87, one should be able to print images from both applications and they should appear the same.

For super color geeks who might ask what was that one worst patch who's dE was 2.81:

92.0/0.0/0.0 20.53/31.36/21.63 92.0/0.0/0.0 22.33/32.28/17.99

First set (92/0/0) is RGB, 2nd value is measured color in Lab. So the big difference is in bStar.

Nice to see the photography side of the site back,

I'll second Jim Meeks, "how does this apply to black and white printing or does it?"

Epson has advanced black and white. How best to use it? My thoughts are to get the best image I can using Silver Effects, and then? Leave it to the printer? How would I soft proof?

I am under the impression that Ctein is a very experienced, accomplished, and successful printer of other's people's photographs as well as his own -- so although there may be merit to the argument that his prints would be even better if he were to use custom profiles, the question on my mind is "better as compared to what?" Better than 90% of the best prints being produced these days? Better than 97.6%? Would any of his clients be able to notice the difference or even care? Is a job done "correctly" always the same as a job done well?

Let's also keep in mind that a disinterest in color spaces, calibration, and custom profiles is not a character defect but rather just a different point of view and a different approach to artistic endeavor. In the end, the work has to speak for itself, profiles or not.

Andrew, you made the point in the video that both images on screen were 16 bit/channel depth. I read that's important, when you are using ProPhoto RGB, because 8 bit (256 values per channel) is too small for all the possible colors in ProPhoto RGB and you can get banding. Do you have any experience with this?

Also in the printer setting, you checked in Photoshop «send 16 bit data» but in the Epson window you didn't check 16 bit. This has no effect?

Thanks for this very good post and video!
Tobias

Tobias, I've never seen any difference sending "16-bit"through the Epson driver. I've never measured any difference either. Note that this isn't even an option for Windows users of Epson drivers. Epson told me that in the future, that might be a different story, hence the 'option' in the driver.

It's a very good idea to use high-bit (what Photoshop calls 16-bit) with wider gamut working spaces although my dear friend and mentor, the late Bruce Fraser was hired by Kodak when they first designed this working space to do testing and he stated at the time that 8-bits per color (24 bit) editing, if not 'severe' was acceptable.

The point to consider is how did your data come to be in ProPhoto RGB? For me, it's a direct encoding of my raw data from Adobe raw processors (Lightroom or ACR). Both products use a variant of ProPhoto RGB for their processing so you're getting ProPhoto gamut at some point in this workflow. You're getting high bit (16-bits) so stick with that. Why not just render from raw into this working space? What's the point of using a smaller working space? If you use another raw converter, this might be a different story. For example, while Apple doesn't tell us, I"m pretty certain it's using Adobe RGB (1998) gamut (primaries) for it's underlying processing, while iPhoto use something very similar to sRGB (it's called "Camera RGB").

In a prefect world, if we had image data that fit just inside sRGB, we'd use that. If it fit nicely within Adobe RGB (1998), we'd use that. But do you really want to go down that path for every image? Or, use the wider gamut working space which in the case of LR/ACR is ProPhoto RGB?

There ARE issues with ProPhoto RGB, if there were one perfect RGB working space, we'd just use that. There are 'colors' (quotes on purpose) or a better term, "Device Values**" that ProPhoto RGB defines we can't see. No display or printer will EVER take full advantage of these device values; we can't see them. You can be editing colors you can't see (although there are ways around this if you're careful).

** http://digitaldog.net/files/ColorNumbersColorGamut.pdf

>>Epson has advanced black and white. How best to use it?

That's a black box, you can't use printer profiles, you can't soft proof. The idea behind ABW is:

The driver does the color to B&W for you, based on the limited controls (It can do a very nice job).

The driver uses less inks for 'better' gray balance and the print is more "archival".

IF you do any conversions on your own, in Photoshop or a 3rd party product and then send it though ABW, it's going to alter your work, process that data as if you just sent it a color image. So it seems a bit pointless to do this.

IF you decide you want to do your own conversions, then you probably want to maintain that appearance so we're back to using an ICC profile and soft proofing. You may not get dead nuts neutral output with an ICC profile over the entire tone scale, that's real difficult to do. So if you say split tone them, or add a color cast or effect, you'll probably be OK with a good ICC profile.

And color gamut isn't an issue here, you could use sRGB.

One more (yes the last) dE report, targets not as dry as I'd like (an hour) but this time, I Assign Adobe RGB (1998) to patches for each application. Also Glossy paper instead of Luster. The numbers are better and I think there should be no debate that you can print your images in either Lightroom or Photoshop and get the same results:

--------------------------------------------------

dE Report
Number of Samples: 1728
Delta-E Formula dE2000
Overall - (1728 colors)
--------------------------------------------------
Average dE: 0.31
Max dE: 1.06
Min dE: 0.02
StdDev dE: 0.18

Best 90% - (1554 colors)
--------------------------------------------------
Average dE: 0.27
Max dE: 0.55
Min dE: 0.02
StdDev dE: 0.13

Worst 10% - (174 colors)
--------------------------------------------------
Average dE: 0.67
Max dE: 1.06
Min dE: 0.55
StdDev dE: 0.11

--------------------------------------------------

Thank you Andrew and Dave for your more than thorough answers and tests. I did my own little test with my 2880 on some Epson Exhibition Fiber paper and could see no difference between PS and LR using the paper profile that a third party supplies for Epson. I did see a difference between the print and my screen... the blue sky showed more purple in both prints. After watching Andrew's video I'm guessed that's because some colors were out out my iMac's screen gamut but show up in print. Either that or I need an up-to-date profile customized to my actual printer output.... whew!

Finally, I did a test print of the same shot using "printer manages color" and AdobeRGB. The sky shows a nice blue much closer to what I see on the screen, and when I compare them side by side, it seems the printer managed to remove a slight color cast (magenta? hard to tell). So... I might actually stick to Ctein's advice until I get my profile sorted.


[Ctein replies: Cyan-blues (such as sky colors) that print out too purple are an indication of an inadequate profile. One of my test images, which I also use in the profile-making process, has just such colors in it. It's a key test for me of how close a profile is to getting things right.]

I'm just typing this with one thumb on an iPhone so I won't going to the gory detail but if you're using a Canon printer they come with a plug-in for Photoshop that behaves completely differently from their printer driver that works with everything else, more like an R I P actually. The plug-in is vastly better then just using the printer driver with white room so I always end up printing in Photoshop if only to be able to use the plug-in.

Talking about printer profiles and calibration without broaching the subject of rendering intent seems more confusing than actually useful to me.

Some of the color spaces have some pretty non obvious limitations, for instance, for instance 16bit HSV has 4,294,967,296 different ways to represent black.

I object to "sub-optimal output" in principle. But I've done enough with inkjet (and darkroom) printing to know that the difference between 95% and 100% can be days or weeks of work. Also that defining "100%" depends a LOT on my mood of the moment.

I'm pretty sure I've never made an exactly absolutely perfect print. And if I thought I had, other people would disagree.


[Ctein replies: Yeah, in general, but in this case we're not talking about trading sweat for quality. It's just about which buttons you push and tabs you click on when you're printing. If one approach gets you better quality than another for no additional effort, seems to me that's a no-brainer win. ]

Glenn Brown: "Nice to see the photography side of the site back."

{sarcasm} Where? {/sarcasm}

Wait, Dave Polaschek said this in passing:
Picking "Printer Manages Colors" on Windows guarantees that your color data will be converted to sRGB. We could adopt XPS printing to get around that, but there are many new and exciting bugs waiting to be found by the first application to head down that path.

Does this really mean that you can only print from a sRGB color space in Windows??? (Using PS or Lightroom)

Thank you for the interesting discussions.

What are 'optimal' profiles? Are the profiles from SpyderPrint (which I have) or ColorMunki deemed good enough in terms of this discussion? Or are we talking about profiles from much higher end equipment? Are we better to send away for such profiles?


[Ctein replies: If you're working in Adobe RGB space, then ColorMunki is definitely capable of producing profiles as good as any commercial ones I've obtained to date (Andrew is going to put that to the test). If you're working in ProPhoto RGB, it won't cut it. I just ran the test, creating a new profile using Andrew's Gamut Test File, and the results were just marginally better than my previous profiles. That is, they still had serious problems dealing with the most extreme tones and colors the P800 printer could reproduce from a ProPhoto RGB file.

It could be assumptions built into the ColorMunki profiling software; it could be something as simple as the puck being unable to read densities that high in print. Whatever the reason, ColorMunki isn't up to the task.

I'm afraid I can't tell you anything about SpyderPrint, as I don't have one.]

Dave Polaschek said: "Picking "Printer Manages Colors" on Windows guarantees that your color data will be converted to sRGB."

Like Trecento, I'd like more info on this. Which piece of software does the conversion and why? Normally you'd set "Printer manages colours" for printing ABW, which is what Ron Martinsen recommends, but if this is correct as stated, there's always a conversion to sRGB in there somewhere. Does this really always happen? Or only in certain circumstances?

Brian Stewart said: Normally you'd set "Printer manages colours" for printing ABW, which is what Ron Martinsen recommends, but if this is correct as stated, there's always a conversion to sRGB in there somewhere. Does this really always happen? Or only in certain circumstances?

This always happens. When using "Printer Manages Colors" on Windows, Windows ICM 2.0 assumes that any color data passed to it is in sRGB, so Photoshop converts the colors to sRGB (so users don't get wrong colors when printing). We can't know whether the user has picked ABW or not (the driver-specific print settings are a black box to us at this point), and we must provide sRGB to the OS in this case.

What Dave wrote is pretty clear to me, IF you use Printer Manages Color (not Application Manages Color which I advocate, with ICC Profiles), YES, you clip colors to sRGB on Windows. That's bad gang! Just test it. You have access to the Gamut Test File or use your own images in a wider gamut RGB working space, print both ways. Here's a color reference image in Adobe RGB (1998) to test as well (the Gamut Test File is better for this test however):
http://www.digitaldog.net/files/2014PrinterTestFileFlat.tif.zip

FWIW, clipping color and gamut compression are not the same! Clipping to sRGB is like a sex change operation using a pair of rusty, dull scissors!

All this is still basically subjective, as the comparisons between Ctein, Rodney, et al do not control for major variables between their setups, procedures or output measures. As a practical matter, each printer needs to decide for him/her self what works best. When all is said and done, the final test, and the one that counts is: How do you (and your customer if you are a pro) like the results.


[Ctein replies: I would disagree. How much color accuracy you need is a personal choice, but color accuracy is substantially an objective metric. You don't know what variables we are controlling for (which is a lot more than you believe), because we're not providing them in tedious detail. You would err to assume absence of evidence is evidence of absence. This isn't a scientific publication, it's a popular one.]

I agree with a comment that Jonathan Sachs made on a bulletin board: "I don't really recommend using PhotoRGB as a working color space - it is too wide a color space and wastes much of the gamut specifying unachievable colors and does not dedicate enough space to represent the more common, less saturated colors accurately."

My working space is Adobe RGB in Raw Therapee and Picture Window Pro. I print using an ICC profile. Occasionally I have tried PhotoRGB as the working space in those programs. Without describing the problems, I have as an unproved hypothesis that conversions to the monitor (a NEC wide gamut display) introduce as many distortions as the use of PhotoRGB is supposed to avoid.

Dave Polaschek said: "This always happens. When using "Printer Manages Colors" on Windows, Windows ICM 2.0 assumes that any color data passed to it is in sRGB, so Photoshop converts the colors to sRGB (so users don't get wrong colors when printing). We can't know whether the user has picked ABW or not (the driver-specific print settings are a black box to us at this point), and we must provide sRGB to the OS in this case."

Thank you Dave and Andrew for making that crystal clear. Is this documented anywhere? Rest assured that I almost always use application manages colour with ICC profiles for colour work, but there is the question of ABW. Also, for some simple print jobs I sometimes want to send an image in AdobeRGB to the printer and use the ICM option in the driver. I suppose that ACPU is the the solution in both cases, or Qimage. Does this also apply to applications like InDesign?

And here I was thinking that Mac users had it bad with the difficulties of printing colour unmanaged files since 10.6.8. Seems they're not alone.

Ctein,
OK you are controlling most/all variables. Not detailing them is understandable in this forum, but if you are making objective measurements of differences in color rendition, please indicate how-spectrograph, or ??. In practical terms, though, few have the equipment for good color comparisons if the difference is small, so most of us will have to make the decision based on perceptual "measures" -what we, or the customer likes best.

>>All this is still basically subjective, as the comparisons between Ctein, Rodney, et al do not control for major variables between their setups, procedures or output measures

It can be subjective or colorimetric (the only way we should define color accuracy) or both.

My video provides a file and testing methodology anyone can try on their end, with their equipment WITH ProPhoto RGB data from raw (or in the case of Bill's balls, produced in Lab). This is subjective. You view the prints each way and come to your own conclusions.

Anytime one speaks of color accuracy NOT pleasing color, we need a way to define accuracy numerically (otherwise, we're back to subjectivity). That's where colorimetry comes to play. But, colorimetry is about color perception. It is not about color appearance. It's important to recognize the differences and when to use each!

IS the profile colorimetrically (numerically) accurate? That can be tested but it's geeky.

In terms of all color management, and current color science not based color appearance, all the CMS knows about is a single colored pixel. Color management doesn't know anything about color in context! That's why we need to view millions of solid pixels such they appear like an image. And the reason why LOOKING at something (subjective analysis of pleasing color) is more valid than measuring it is (Colorimetry which provides a metric of accuracy) is because measurement is about comparing solid colors. Color appearance is about evaluating images and color in context which measurement devices cannot do.

I don't really care about color accuracy, for one thing it's hard to even define much less achieve, and more often than not it looks bad anyway.

For instance imagine a portrait of a woman sitting near a window in a bar during daylight with a neon Budweiser sign in the window, a mix of fluorescent and incandescent illuminating the interior background, with a TV set over the bar. You might want realistic color, or plausible color, or creative color, but probably not accurate color whatever that might be.
Ok for extra credit, is the client Budweiser or the woman's record label?
Just be thankful it's not rolling rock.

On the other hand, I do care about consistent, repeatable, and predictable color quite a lot.

I re-read the comments and get the distinct impression that we are variously discussing apples, oranges, tennis balls, and bicycles.

Perhaps someone should define the terms here.

When I think of accurate color reproduction I am generally thinking of a reflective print that looks the (reflective) object when placed next to it. Art reproduction and product photography for example.
Anything else is some sort of subjective idealization. Since most photography takes place in some sort of uneven uncontrolled mix of light sources where the goal is to render the scene , and not some idealized object in the scene, it falls into the realm of subjective idealization.

In cases where photography is a way of describing a group of three dimensional objects each of which is some combination of both a light reflector and a light source, and I think the majority of photographs are, then what constitutes accurate color?

For those who are thinking of accurate color as describing a situation where a reflective print matches transmissive / lightsource image like a monitor or a color transparency, well that just isn't going to happen outside very contrived situations if ever.

So, what are we talking about when we talk about accurate color? I would argue that every definition is in some way wrong as applied to general photography, but I'd settle for agreeing to the same wrong definition, or some very limited definition.

[Ctein replies: Hugh (and others), the scope of the discussion here is limited. The function of monitor profiles is to ensure that your monitor displays as accurate a rendition of what's in your image file as possible. The function of printer profiles is to ensure that those colors are printed out as accurately as possible.

It has nothing to do with how well the color in those files compares to the original scene or what your artistic intent is. Those are entirely separate matters. How you get from a real-world scene to an image that looks the way you want it to look, well… That's your problem. All that color management is concerned with is trying to ensure fidelity after that point.]

If display drivers use 8-bits per color and printer drivers use the same how can any image in Profoto or Adobe RGB be displayed or printed without clipping? Why wouldn't sRGB make more sense? Just asking because this has puzzled me for some time and I have never seen a good explanation.


[Ctein replies: Paul, it's easiest for me to explain this if I start with exposure range. There are people who think the exposure range of a camera is controlled by bit depth. It's not. Bit depth and exposure range are two different aspects of image quality. One does not convert to the other. You can think of the total exposure range as being a staircase. How many stories up the staircase runs is how long the exposure range is. How many steps are on the staircase is the bit depth.

Exposure runs on a line. We can extend that metaphor to area (or even volume) for color spaces. What your bits do is parcel up that color acreage into subplots. The more bits you have, the smaller the plots; in other words, the more distinct colors you can display or print. But your total acreage doesn't much change with bit depth.

The reason for working with 16-bit files when you're working in a very large color space like ProPhoto (or Wide Gamut) RGB is that each subplot becomes so large that you're likely to see contouring at the transitions between adjacent colors. There's a lot more area, so you need a lot finer divisions.

The reason this isn't a problem for displays or printers is that they can't render anywhere near the entire ProPhoto RGB color space. Most displays are close to sRGB, the very best displays and printers render a little larger than Adobe RGB. The divisions produced by 8-bits-per-channel color are fine enough that you don't see contouring.

At some point we may have displays and/or printers that can portray a large enough chunk of the color space that it will be important to have more bits per channel. So far, though, we have no real-world examples where sending 16-bit color to printers that have a 16-bit-compatible driver produces a visible improvement in the print.

I hope that clears everything up.]

>>If display drivers use 8-bits per color and printer drivers use the same how can any image in Profoto or Adobe RGB be displayed or printed without clipping?

Bit depth has nothing to do with color gamut:

http://digitaldog.net/files/ColorNumbersColorGamut.pdf

When you work with 24-bit images, all "color" and tone is defined in three 8-bit color channels. When you work with wide gamut working spaces, the same bits need to be spread farther apart over the entire color space. Consider this spreading of a finite number of bits as follows: Imagine you have a half-inflated balloon that has 16.7 million dots evenly spaced over its surface. Now you blow up the balloon to twice its original size. Each dot is spread farther apart. When you work with 8-bit-per-channel files, you create this effect when you encode the bits into a progressively larger gamut working spaces. In such situations, it is possible that editing images will produce banding(aliasing). For this reason, should you decide to use a wide gamut working space—for example,something wider than Adobe RGB (1998)—you should attempt to encode the data in 16-bit color.Many capture devices produce more than 8-bits per color and allow you to retain this extra data
to use in Photoshop. While the file size will be twice as big and image processing will take longer,you can’t be too careful with your data. You may also wish to use 16-bit data with smaller gamut color spaces.

http://www.adobe.com/digitalimag/pdfs/phscs2ip_colspace.pdf

>>So, what are we talking about when we talk about accurate color?

I think the term accurate color is misused here. The term you're looking for is pleasing color.

Accurate color can only be evaluated with two color values or samples: Reference (what the colors should be, defined in Lab) and what colors are produced (again in Lab). Then we can use a metric called deltaE to exactly define color accuracy. Lab was designed specifically for this purpose! To tell us how different one color is from the other. So if you want to know how accurate your printer profile is, you can send a number of colors through it, to the printer and measure the results. You can compare what you just measured to the reference (what the values should be), now you can state the output profile is or isn't accurate by X number of deltaE.

A color match isn't necessarily one where two sets of values match however! There are a slew of cases where color value in isn't numerically anything like color value out and you can still produce a match or what could be considered pleasing color.

Accurate color has no place in this discussion for another reason. The accuracy doesn't take image content or colors in context into account. It only looks at two solid colors (pixels, color patches on a print, color patches on-screen). There's no understanding of color in context, that's something ICC color management can't provide but humans can by looking at the output.

The comments to this entry are closed.