Nerdy Tech-Talk: Some Thoughts About the Adobe RGB Colorspace

16May17

My last couple of blogs seem to have avoided nerdy technical concepts, so I think it’s time to geek out with some tech talk. What follows is an edited excerpt from some comments I was asked to make on a “how-to” document being produced by the British Library’s Endangered Archives Project. They had produced a photographic guide for the field digitization of archival documents and wanted some feedback on its technical recommendations. One of the topics discussed in the document was the subject of Red-Green-Blue (RGB) colorspaces, specifically the differences between sRGB, which is commonly used in many digital processes, and Adobe RGB, which is often touted as the more “professional” choice for folks wanting a larger colorspace. I felt that a discussion of Adobe RGB’s advantages and potential pitfalls might be helpful. Note that in all discussions here we are talking about 8-bit-per-channel color, where intensities range from 0 (none) to 255 (all), allowing for a total of 256 levels (2 to the eighth power). Read on.

The issue of what colorspace should be used for imaging is a surefire way to start a debate among photographers and image curators. As often is the case, there is no single “right” position, but there are things to consider when making a decision on this.

Back in 1998, users of Adobe Photoshop were offered the company’s newest release of the software, version 5.0, which advertised a new feature called “color management.” Professionals who had been using Photoshop for years probably figured that they were already “managing color,” but this promised to be an entirely new technology. One thing that the user needed to do before these new features would be of use was to choose a “working RGB colorspace.”

I was working as a commercial photographer for a prepress service bureau at the time, and I can attest to the fact that most users, even experienced retouchers, didn’t really know at the time what “working colorspace” meant. We worked in “RGB,” right? And, after things were worked on, the RGB file was converted to Cyan, Magenta, Yellow and Black, (CMYK) for lithographic printing (we used a standalone $3000 piece of software just for this purpose). The files were then “scattered” to film by the imagesetter and proofed on Agfa Pressmatch material. If something needed further work, you went back, did more corrections, and ran the proofing process again.

For a color-managed workflow to function, however, the user had to choose a default RGB colorspace to work in—this would work in conjunction with the color profiles created for monitors, capture devices and output devices to create a situation where color would remain consistent throughout the capture, editing and output processes.

Photoshop offered a number of choices for a working RGB colorspace. The most familiar to us was Apple RGB which was modeled after the display characteristics of a high-end Apple Trinitron (tube type) computer monitor, which was the type of display we all used at that time. This colorspace has a white point of “D50” (5000 degrees Kelvin) and a native gamma (power response curve exponent) of 1.8, which was typical for Macintosh computers at the time. There are differing stories as to why these specs were chosen, but the most common explanation was that it best simulated what things would look like when printed.

If one wanted to use a colorspace that could encompass a slightly wider gamut (range) of realizable colors, one could choose ColorMatch RGB. This mimicked the display characteristics of the slightly pricier Radius PressView computer monitor, which was a high-end aftermarket display for Macs at that time. It too had a gamma of 1.8 and a white point of D50, but the Red, Green and Blue primaries were farther “out there” in the available visible color gamut than those of Apple RGB.

sRGB was another choice, a colorspace proposed and described by Microsoft and Hewlett Packard in 1996 and implemented in Windows 98. It had RGB primaries that were similar to Apple RGB but with a gamma of 2.2 and a slightly “cooler” white point of D65 (6500 degrees Kelvin). Being Mac snobs, we weren’t about to use anything conjured up by Microsoft for our precious color retouching needs, thank you very much!

And then there was this choice: SMPTE 240M. The acronym stands for the Society of Motion Picture and Television Engineers. It had a much larger color gamut than the previous three, a gamma of 2.2 and a white point of D65.

If you looked at a CIE color diagram, that “sail-shaped” graphic with the colors and triangles indicating various color spaces, it was obvious that SMPTE 240M was better at encompassing the total colorspace of typical high-end lithographic printing than any of the aforementioned ones. User and folks writing about Photoshop starting experimenting with the cryptically-named colorspace because of this.

Bruce Fraser, a famous author of books on Photoshop, called SMPTE 240M the “aggressive choice” when working with 8-bit-per-channel RGB workflows. Most of Photoshop’s advanced functions did not work on 16-bit images at the time, so there was a danger in working with a colorspace this large with only 256 available intensity levels for Red, Green and Blue. Since these numeric “steps” stretched out to describe such a large space, severe edits could cause posterization or banding in images, especially in the darker areas. Fraser even went so far as to propose his own custom RGB colorspace that fell between SMPTE 240M and ColorMatch RGB. He dubbed it Bruce RGB. It didn’t catch on, though it was a good idea for the 8-bit workflows we were using.

Most of us in the professional retouch field settled on ColorMatch RGB, which more closely matched our perception of what a quality Mac display should look like and had “better numbers” in terms of RGB primaries than Apple RGB or sRGB.

Adobe Systems was also looking at SMPTE 240M, and, if you believe the Wikipedia article on Adobe RGB, wanted to implement it as a standard because of its ability to encompass all the realizable colors produced by CMYK printing inks (remember that at this time, the printing industry was the primary destination of quality digital imagery, not the internet). They, according to Wikipedia, misinterpreted the “idealized” RGB primaries described by SMPTE 240M as the actual ones, and made an additional mistake transcribing the Red primary coordinates.

If this sounds ridiculous, consider that NASA lost the Mars Climate Orbiter in 1999 because one of its teams provided calculations in English units and another provided calculations in metric.

This became Adobe RGB

It’s nearly 20 years later and Adobe RGB is one of the standard choices for RGB work and is considered the one more suited to professional use, although my opinion is that this never would have happened had not the name “Adobe” been assigned to it.

CIE diagram.

CIE Spectral Diagram with RGB Colorspaces. This is a two-dimensional representation of a three-dimensional space, but it serves to show the difference between Adobe RGB and other common colorspaces used by imaging technicians. The outer border of this sail-shaped diagram is knows as the spectral locus.

Here’s the famous “sail-shaped” diagram that we’ve all seen. The important thing to know is that the farther from the middle (the white point) the more saturated or “pure” a given color is. So, looking at this, Adobe RGB is obviously superior, right?

Technically, yes. A given color number in Adobe RGB, such as “Red 255” would be a redder red than Red 255 in sRGB. It’s also the same for the other two primary colors. In fact, sRGB Red 255 (the highest number in 8-bit color) is the same actual color in Adobe RGB as Red 219!

Different examples might involve all three colors. Maximum Green 255 (other color values at zero) in sRGB converts to the Adobe RGB triplet of:

R 144
G 255
B 60

Here, we see that “full pure green” in sRGB is the same color in Adobe RGB as its maximum green plus contamination from the other two colors. Therefore, we know that “max green” in sRGB is definitely not the purest green that Adobe RGB can represent.

In fact, Adobe RGB can represent colors far into the Cyan and Green region of the spectral locus that far exceeds what sRGB can reach. This is why Adobe RGB can encode RGB color values for pure CYAN, one of the few CMYK colors that is slightly outside the sRGB color space, which was one of Adobe’s main goals in adopting this space.

In Blue, the difference is less severe. sRGB Blue 255 (other colors at zero) translates to Blue 250 (no other colors) in Adobe RGB.

For a color management system to work, an image file needs to have an embedded International Color Consortium (ICC) color profile along with the numeric pixel color value encoding. This profile is a small bit of metadata that indicates the colorspace characteristics. A “profile-aware” system can read the numeric pixel information, refer to the embedded profile, and accurately render the true visual color that a given RGB triplet represents based on the color space specified.

Challenges of Working with Adobe RGB

Editing

When sRGB was introduced, its color primary limits were representational of the color limits of high-end monitors and eventually became the model for HDTV (in the guise of a standard called Rec 709). Therefore, Red 255, Green 0, and Blue 0 in the sRGB colorspace is the reddest red that a monitor or HDTV can display. Using the previous example, though, we see that this means that when Adobe RGB reaches Red 219, it’s at the limit for what the monitor can display for Red (it’s the equivalent of sRGB Red 255). Therefore, how would we view the Red values above 219 in an Adobe RGB image seen on a monitor?

The short answer is you don’t, not yet anyway on the vast majority of computer monitors and flat-screen TVs that exist today. There are new computer monitors being developed that claim to have the capability of displaying the entire Adobe RGB color gamut, but they’re not common and will be an expensive option for graphics professionals for some time.

You can try this in Photoshop yourself. Change your working color space to Adobe RGB and open a new RGB file (any size). Change the foreground color to:

R 255
G 0
B 0

Fill the canvas with this color; it will be an intense “candy apple” red. Now, change the foreground color to:

R 219
G 0
B 0

Create a rectangular selection in the middle of your intense red canvas and fill it with this color, which is not pure red. Now drop the selection so the “marquee box” is gone. On my (inexpensive) monitor, there’s no difference in tone and you can’t see the inner box. Adobe RGB Red 219 hits the Red limit on the monitor and Red 255 is even redder than that, but the monitor can’t show it because it’s already at the red limit.

So how do you perform color editing on a monitor in a colorspace that has colors that exceed what the monitor can show you? You go into the color settings in Photoshop and click on the “more options” button to show the “advanced controls.” There, you can find a checkbox marked “Desaturate Monitor Colors By:” with a percentage box next to it (it’s set by default to 20%). Click this, and suddenly your inner red box appears within the outer box.

Advanced controls

Advanced controls for color settings in Photoshop. Here, the setting to desaturate monitor colors has been turned on, allowing the user to see color relationships between saturated colors in colorspaces that exceed monitor capabilities.

The helpful hint displayed by Photoshop indicates that this setting is recommended for “experts only” and that the screen display will no longer match the printed output. However, it’s the only way to see relationships between saturated colors in Adobe RGB that a typical monitor cannot reveal. For me, this is a nerve-wracking way to work because you have to switch back to normal viewing to see what your image is really going to end up looking like overall.

Now, unclick this desaturation control (the inner box again disappears). Now, do an “assign profile” to sRGB (I use relative colorimetric rendering intent with black point compensation). Now, the outer box is still the fullest red the monitor can produce, but the center box desaturates to what Red 219 actually looks like in sRGB.

Display issues

The phenomenon illustrated by the previous exercise brings up another potential problem. If Adobe RGB images are viewed using software that is not color-managed (that is, it doesn’t recognize or respect the embedded ICC profile),

OR

for some reason, Adobe RGB images are saved without the profile embedded,

OR

if for any reason the ICC profile is removed from images that are supposed to be in Adobe RGB,

THEN

the images viewed on a monitor will appear desaturated, not possessing visual fidelity to the original object that was scanned or photographed.

This is due to all the numeric stuff described previously. If an Adobe RGB “Red 219” pixel comes into a system that doesn’t read the profile and accurately equate it with sRGB Red 255 (as it should, producing the reddest red the monitor can make), it will simply display it as Red 219, which on a monitor is a desaturated red. All the other colors follow suit.

This problem was widespread only a few years ago, when most web browsers did not read and implement embedded ICC profiles in images. Anything that was intended for Internet use had to be converted to sRGB unless one was willing to accept the desaturated look that would result from Adobe RGB images displayed without the proper color-matching to sRGB. Today, most web browsers will respect profile information embedded in image files. I’ve tested Windows Explorer, Chrome, and Firefox in Windows and Chrome, Firefox and Safari on the Mac and they all seem to work. I’ve read that the version of Apple Safari for IOS (iPhones and iPads) also works. I’ve also read that the default Android browser does not, and that various older browsers in multiple platforms may not.

So if you’re producing images for the web and you work in Adobe RGB, you don’t know for sure if the user will be seeing the images rendered properly (ICC profile respected and implemented) or slightly desaturated (profile ignored and RGB numbers simply interpreted as monitor RGB). There are no guarantees, though we can hope that users are using up-to-date browsers.

2017 Adobe RGB Red Test

2017 sRGB Red Test

Test Images for browsers. Both square have four distinct “rings” of red intensity, with the outer ring being Red 255 and the middle being Red 220. The top square is in Adobe RGB; the bottom in sRGB. If your browser is doing color management properly, the top square should look like one solid color, unless you have a very high-quality monitor, in which case you might see a bit of the center portion.

Other image viewing applications are likewise “hit and miss” regarding the proper reading of embedded profiles in images. IrfanView, a popular PC image viewing program, does not read color profiles without installing a plugin. FastStone Image Viewer also does not unless the user deliberately enables its color management feature. Again, what images end up looking like is often out of the control of the producer of the images and at the mercy of whether the user has set the software up correctly.

Imaging technicians must at the very least make sure that if they are working in Adobe RGB, the images they create have to contain the embedded profile. If not, they are unwittingly working, in practical terms, in a desaturated sRGB colorspace and will produce inaccurate-looking results.

Digital dSlR cameras often give the user a choice between Adobe RGB and sRGB. For some time, certain Nikon dSLR models had a bug in their firmware whereby the system would not embed the proper profile in their RAW files (NEF is Nikon’s RAW file type), producing the same sort of desaturated result described above if Adobe RGB was chosen as the working space. In choosing to use Adobe RGB as a digital camera device profile, I would run some tests to make sure that the resulting RAW files contain the proper profile and that the resulting images look normal and not desaturated. This is where photographing something like a Macbeth Color Checker would be quite useful for evaluation.

If images are destined for certain types of display systems, and it is known that the system isn’t “profile aware,” the only choice for users with Adobe RGB images is to create sRGB derivatives for use with that particular system.

At HMML, we have our dSLR cameras set for sRGB. This simplifies our work considerably, as we use a “RAW+JPEG” setting in the camera where a high quality JPEG is created from the RAW on the fly as photography proceeds. This saves us the considerable time it takes to convert RAWs to JPEGs or TIFFs, and also the time it takes to create derivatives in sRGB from RAWs created in Adobe RGB. These time savings can be considerable; we recently converted the RAW NEF images for 220 manuscripts into color-corrected derivative JPEGs for the web and it took over 20 hours.

Some would claim that this isn’t the way to go and that we’re somehow lacking color in our images because of this decision. To this I would point out that having an image in the Adobe RGB colorspace does not by itself make it more “colorful.” It simply means that the colorspace parameters allow for a larger color gamut to be described by the range of RGB triplets available. In the sort of manuscript photography that we do, it is highly unlikely that any subject matter being photographed contains colors outside of the sRGB color space, so we’re not actually losing any information.

In fact, you have to look for specific types of subject matter to find colors in the real world that inhabit the area of the spectral locus covered by Adobe RGB and not covered by sRGB. Lights with pure color filtration might be one thing. Pure cyan lithographic ink on top-quality paper is another; some of the pure inks (no other colors to contaminate them) used by inkjet printers on high quality paper might be as well. But HMML’s not photographing swatches of pure inkjet ink on photo-quality paper; we’re generally photographing slightly brown paper with brownish-black ink, and an image like that rendered in Adobe RGB or sRGB will look the same; those colors will simply be described using different RGB triplets if one checks the numbers between the two versions.

In working with 8-bit color, where I have a total of 16,777,216 (256 to the third power) possible RGB triplets at my disposal, I don’t like the idea of wasting a bunch of them on color values I will never have to use based on the subject matter I’m recording. This is what my former boss from the service bureau calls, “dog whistle territory.” Lest anyone think that 16.7 million possible RGB code points is a lot, it should be noted that using the mathematics of color science and human visual perception, “seamless” RGB tonal gradations should actually require a minimum of 463 levels per color instead of 256, but most subject matter is forgiving enough to not display posterization or banding at 8 bits. If human visual perception didn’t have a gamma of around .4 (nearly the reverse of the curve described by gamma 2.2), 8-bit digital color wouldn’t be feasible at all.

But that’s a topic for another day.

Advertisements


One Response to “Nerdy Tech-Talk: Some Thoughts About the Adobe RGB Colorspace”

  1. 1 Dean Severson

    Wow, nicely written. Explaining things I never really understood before.


Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s


%d bloggers like this: