This page is devoted to data and descriptions for individual color puzzles and adventures, each centered on a particular color that presents interesting color correction problems.

The Mystery of the Blue Glass,
or Profile-Schmofile!

>>>

Odds are, there are intelligent aliens out there somewhere in the universe, and that they are at least as busy as we are, making cobalt blue glass, the same as us. It's a great saturated and pure color, and a permanent because it relies on the optical properties of cobalt oxide. If the alien species I mentioned are sensitive to infra red the glass will be a nice shade of purple to them. This is because it also reflects infra red light in addition to pure blue.

Even if you don't believe in aliens, you may be surprised to learn that these "alien species" live among us, masquerading as digital cameras. Furthermore, we have created software that often finds it convenient to mess with cobalt blue, making it darker and/or more purple. And third, when we rely on ink to create our images, pure bright blue is the most difficult color of all to get right.

One fascinating aspect of the color of this glass is that it is easy to capture on a digital camera, easy to see well on a monitor, and almost impossible to print on a printer.

As a matter of fact, unless your printer is set up to a gnat's eyelash, the images shown below will look completely different on your printer, taking on an opaque appearance that is not at all like our friend cobalt blue. Try it and see!

sRGB	Adobe RGB	ProPhoto RGB
Figure 1. The three images above were converted from the same raw file using Adobe Camera Raw 3.1, saved using the indicated profiles, then converted to sRGB for display on this web page. The original Tiff images may be downloaded below for direct viewing in PhotoShop

For example, what happens when we print each of these on an Epson 1290, using profiles from Epson? Since this is a web page, I will use PhotoShop's soft proofing feature on the original tiff images for each of the three profiles represented above, I have used the profile for the 1290, and used screen captures to simulate what that printout might look like, below.

sRGB soft proofed to Epson 1290 MP-H	| Adobe RGB soft proofed to Epson 1290 MP-H	ProPhoto soft proofed to Epson 1290 MP-H ... Ew!
Figure 2. These images are screen captures of soft proofs done using perceptual intent, with Black Point Compensation selected, and Black Ink and Paper White not selected.

Looks pretty bad, eh? We would expect to lose saturation with a deep blue object like this, while retaining transparency, as seen above. SRgb and Adobe RGB look about the same. Surprisingly, the glassy look is almost dead in the ProPhoto RGB soft proof. Check out the upper right corner, where most the transparent area appears darker. Is this a problem with ProPhoto RGB?. No, it turns out its the printer profile, as we'll see below.

Let's repeat the above soft proof, only this time using a canned profile that Epson ships with its Color Stylus 1270 printer, an older cousin of the 1290.

sRGB soft proofed to Epson 1270	Adobe RGB soft proofed to Epson 1270	ProPhoto soft proofed to Epson 1270
Figure 3. Much better now!

What a relief! This printer uses the same ink and nozzle technology as the 1290, and the glass looks much better. It has lost a little saturation, as expected when printing such a blue object using CMYK technology. In spite of that the glass still looks relatively clear, and has not taken on any of the opaque appearance of the previous example. (BTW -these soft proofs are very optimistic portrayals of what the actual output looks like. If your printouts look as good as this, stand up and give yourself a salute - you either have an excellent printer, or you're living on a different planet than I am, and either of these is noteworthy.)

Now the rubber hits the road. When two profiles for very similar printers give different results, it's time to trot out Curvemeister's nifty profile plotter action, available as a free download from Curvemeister. Here are two plots, one of the 1290 profile, and another of the 1270 profile. As you can see, they are completely different!

Plot of 1290 profile. Ouch!!!	Plot of 1270 profile
Figure 4. These graphs show what each profile does to a gray ramp, with input values increasing along the horizontal axis.

OK, smoking gun time. If this were a murder mystery, Charlie Chan would be introducing number one son, and the 1290 profile would be gnashing his teeth with unrepentant guilt. Clearly, red, the arch enemy of blue, is being added to the shadows by the 1290 profile. Very bad for glass, particularly blue glass! To add insult to injury, the amount of blue in very dark colors is being reduced. Even without these problems, the jaggy lines in the red and green channels will guarantee some nasty banding for images that require smooth transitions in the shadows.

The 1270 canned profile, on the other hand, is boosting the shadows a little bit, and changing the overall brightness. In keeping with it's generic nature, this profile is at least not monkeying with the different color channels.

As a final experiment, let's do a profile post-mortem by running the 1290 profile on a 256 step gradient, and check each channel for banding.

Red channel after applying 1290 profile	Green channel after applying 1290 profile	Blue channel after applying 1290 profile
Figure 5. The red channel is considerably lighter in the shadows, and there is banding in the Green and Blue channels.

Notice that the red channel is considerably brighter in the shadows - this will give a red color cast to the darkest areas of any image. The green channel is darker, and shows noticeable banding in several areas, particularly in the shadows and midtones. The blue channel is even worse - very dark in the shadows, and with two prominent bands about 20% from the left. It's no surprise that this profile ruins dark blues, and I think it is likely that there are problems with dark shadows of almost any color.

One loose end remains. Why did the ProPhoto RGB image fare so poorly in the second set of images, with the 1290 profile? My guess is that the ProPhoto RGB colors just happened to overlap one or more of these bands in Lab space- it could just as likely have happened with sRGB or Adobe RGB. If you want to scope this out, eyedropper the color in question, fill an area, and see what happens to them when you soft proof using the 1290 profile.

The 1290 profile in question was obtained from the Epson web site. These are "use at your own risk" profiles, and are distributed with no guarantees, and the understanding that they are to be used at the discretion of the person downloading them.

Moral of story - run any new profile on a gradient before trusting it with your ink and paper. Even better, run the profile plotter and have a look at what's happening to your colors. - Mike Russell

The Players
A capture of sunlit cobalt glass, as a Nikon Raw (NEF) file. If you have the Adobe Camera Raw plugin, or another raw file utility that can handle Nikon raw (NEF) files, you may use this file to recreate the tiff files, included for convenience below. BTW - For those who don't have the correct version of Photoshop under Windows XP and would like to sing along, RawShooter is an excellent, free program that will handle a variety of raw files.	DSCN2194.NEF
Converted TIFF versions of the image, made using Adobe Camera Raw, and targeting the ProPhoto RGB, Adobe RGB, and sRGB color spaces. These images may be used to show your own, more accurate, soft proofs. As with all TIFF images, you may need to right click on the link, and specify "Save Target as ..." to save this file to a folder on your computer.	DSCN2194-AdobeRGB.tif DSCN2194-ProPhotoRGB.tif DSCN2194-sRGB.tif
A print profile for an Epson 1290, Matte Heavyweight, that was originally used to try to print the blue glass. If you own PhotoShop, you may use this profile with PhotoShop's Soft Proof feature to get an idea of what the printout looks like. If you have a 1280 or 1290 printer, you may re-create the entire mystery here.	SP1290_MP-H.icm
A clean 255 step gradient, for use in testing profiles for banding in the shadows. As with all TIFF images, you may need to right click on the link, and specify "Save Target as ..." to save this file to a folder on your computer.	Click to download a TIFF image containing a 256 square "clean gradient".