Color Management Myths 21-25
This page is a
For more details see
Reserved ColorWiki Articles
This reserved article originally appeared in CHROMiX ColorNews Issue 16 on Dec 22, 2004.
Click here to see the original in its original context.
Email colornews(at)chromix.com to subscribe to the ColorNews newsletter.
by Steve Upton
Myth #21: There are perceptual rendering intents available when converting from scanner/camera->workspace or workspace->workspace.
First, a bit about non-colorimetric rendering intents and LUTs. Most monitor and RGB working spaces are made using simple Matrix profiles. Matrix profiles contain color measurements for the RGB channels, curves for each RGB channel and a white point tag. They are small, easy to store and totally depend on the device in question behaving very predictably. In the case of CRT monitors and working spaces, this is no problem.
However, when these matrix profiles are used in color conversions, the CMM has little choice of what to do with out-of-gamut colors. There are no instructions in a matrix profile regarding how to treat them. CMMs like the Adobe ACE CMM in Photoshop and other Adobe applications will simply "clip" out of gamut colors to the gamut of the destination. This is the relative colorimetric intent. The perceptual and saturation intents require more detailed instructions on how to treat out-of-gamut (and in-gamut) colors and these instructions are stored in 3D lookup tables (LUTs). So, matrix profiles are simple but can only do colorimetric conversions and LUT profiles can contain the more complicated instructions required by perceptual and saturation rendering.
See the problem yet?
Another point that needs to be made is that under version 2.x of the ICC spec (not the current spec but the one in wide use today) only one rendering intent is required in input profiles. What this means is that scanner and camera profiles are only required to contain one rendering intent's lookup table and that is the colorimetric one. They may contain additional tables but as they are not required, they are often not included.
Getting the picture? No?
In plain English this is what it means:
- when converting from workingspace to working space (say, Adobe RGB to sRGB), the ONLY rendering intent available is colorimetric. Photoshop allows you to select perceptual and saturation but they make NO difference (try it sometime). Photoshop should gray out intents that don't exist in the profiles.. but that's just my opinion....
- when converting from a scanner or camera profile to a working space the ONLY rendering intent available (in many cases) is relative colorimetric. Again, Photoshop will let you think you have others available but they are not contained in the profiles and choosing them will still give you relative colorimetric.
- as a result of these missing intents, you run the risk of clipping out-of-gamut colors AND because many of them are outside of your monitor's gamut you won't notice the problem until you go to print.
Joseph Holmes, a well-known and respected photographer and digital imaging fellow decided years ago that this was not what he wanted and he devised a strategy to get around the problem. His idea was to convert from an input profile to a large working space (and he invented his own, Ektaspace) and then archive the image in this form. When he was ready to go to print he would let the (actually existing) perceptual intent in the print profile shrink the gamut of the file in a pleasing way. This way he would bypass the potentially destructive scanner->smaller workingspace or workingspace->workingspace conversions.
The ICC v4 specification allows for more intents in input profiles but there is no guarantee they will contain perceptual or saturation tables. Over the longer term though we should expect to see these tables in input profiles and we'll have a safer conversion path.
Matrix to matrix conversions, however, will always have this problem so be careful. Select your destination workingspace profile in "Proof Setup" and turn on gamut warning in Photoshop to see which colors in your image run the risk of getting clipped. Then either de-saturate colors by hand or leave your image in the larger working space and let the print profile shrink the gamut when required.
Myth #22: The a, b axis of Lab are red/green and blue/yellow.
The Lab color space as we know it today was created in 1976. The a and b axes are often described by people as the Red and Green axes but this is incorrect. (or at least unclear)
Try this test in Photoshop
- open Photoshop
- open the color picker
- enter the Lab color 100,-128,0 - this is the most negative "a" value
- look at the color that results - it appears cyan-ish
- now enter the Lab color 100,127,0 - this is the most positive "a" value
- look at the color that results - it appears magenta-ish
Sometimes when people write or speak about the Lab color space they describe the "a" axis as the red/green axis. As you can see by the above illustration it is better described as the magenta-ish / cyan-ish axis. These axes were chosen for their positioning relative to human perception NOT to make it easier for writers and speakers to describe them. It's easier to say "red/green" than "magenta-ish/cyan-ish" so many people do it. It's no big deal just don't panic when you discover that they're not actually red and green.
Myth #23: Photoshop's color setting dialog lets you edit an ICC CMYK profile.
This one's from a conversation I had with color management icon Don Hutcheson the other day.
There are two methods for setting up the CMYK working space and CMYK conversions in Photoshop and unfortunately Adobe has not made the distinction very clear so there's room for confusion.
First, the legacy CMYK system that has been around for many many versions is pre-ICC and is available when you choose "Custom CMYK..." in the CMYK popup of the Color Settings dialog. These CMYK setup options allow you to select ink color values, black generation, dot gain and other parameters. Once selected, Photoshop creates a simplified ICC profile that does not contain a paper white value and also only contains the colorimetric rendering intent.
Any other profile selected is a true ICC profile that is used by the Adobe ACE CMM to perform color transformations.
What's the difference? Well, the most important thing is that these are very different methods of setting up CMYK conversions and the older engine, while it has lots of options, produces conversions that are fairly crude and not up to today's standards. Another point is that the settings you see in the Custom CMYK setup area have NOTHING to do with the current or previous ICC profile. It's easy to get drawn into thinking that you can select a modern profile and then alter the profile's settings in the custom area. This is not the case however.
We urge you to use modern ICC profiles for your CMYK setup. if you feel you want to change the separation settings (which is understandable) then generate your own ICC profiles (in a profiling application) using your desired settings and then select them in Photoshop. Photoshop is not a profile editor and the only profiles it can generate are based on old technology and not enough color data to give quality results.
Myth #24: Working space selection in Photoshop affects profile building.
When creating a new print profile, often the first task is to print a profiling target from Photoshop. Whether you are creating an RGB or CMYK print profile the overall technique is the same. Open the target image file without performing any conversion. Then print the file without performing any conversion.
Get the idea? no conversion.
Some people are concerned about their choice of workingspace at the time of target printing. But, if you follow the path of no conversions then the workingspace is irrelevant. The only thing it will affect is the appearance of the target on screen. (for information regarding how confusing this can be, see Myth #11 in ColorNews Issue #13). Myth 11 Otherwise it has no bearing on the profile you build.
There is one case where it may make a difference. ColorVision's scanner-based profiling software makes certain assumptions about outgoing colors when it builds scanner-based print profiles. We don't recommend this method of print profiling as it doesn't measure up to professional quality and it also creates a profile that is dependent on the source color space which is... well.. strange at best, annoying, limiting and confusing at worst.
Myth #25: Device link profiles are limited, hard-coded versions of normal ICC transforms and are therefore redundant, inflexible and are to be avoided.
This one couldn't be further from the truth. But, again, I should start with a little background.
When you do a color conversion in Photoshop or other applications, the two profiles in question (say, US Sheetfed Coated and your inkjet profile) are joined together in RAM to form a single transformation. So, while the conversion theoretically goes from CMYK to Lab through the Sheetfed profile and then Lab to CMYK through your inkjet profile, in actuality the two profiles are joined in RAM and the conversion takes place as a single CMYK to CMYK transformation. This is pretty much transparent to you as a user but it does make the conversion faster and may also increase precision.
A device link profile is what you'd get if you were to save this combined profile to disk. It permanently links the profiles together using the selected rendering intent.
The advantages of this are numerous:
- The CMM that was used to create the transformation contributes to the quality of the resulting device link profile. So you can use a good CMM on your desktop computer to create the link and then bypass the idiosyncrasies of the CMM on your RIP, for example.
- A normal device link does simply encapsulate a typical ICC color transformation including the tendency to reseparate K-only image elements into CMYK image elements. If this is undesirable (which it often is for re-separating CMYK files for alternate press conditions) then some software can filter the device link after it is created so that K-only color remains K-only after conversion. It will usually be tone corrected for the new profile but it will remain "clean". This is often called "preserve black" or "clean black".
- The same "clean" behavior can be applied to the other color channels. As each channel increases to 100% the other inks are minimized, maximizing the purity of the colors.
- An emerging technique is to use device link profiles to automate prepress-style edits. Curve bumps, ink limiting and other techniques can be captured in a device link profile and applied en-masse to large numbers of files in a color server or RIP. This can be a very powerful way of applying traditional channel-style edits in an ICC compliant workflow.
So, while device links are not widely supported (they are still not usable in Photoshop), they are powerful and can solve challenging color and production problems when needed. Keep an eye on technology that makes use of device link profiles. We have only seen the beginning of what they can do.
- EDITOR'S NOTE: Since the release of CS4, Photoshop does handle the conversion of images to device link profiles.
Thanks for reading,
For more Color Management Myths, see Color_Management_Myths_26-28