Color Management: Scanner and Digital Camera Profiling
In the world of color management, ICC profiles are used both to define what exact colors are meant by RGB image values and to describe how the various devices used to process those images respond to color. While standard color spaces such as Adobe RGB and sRGB are generally used for the former, profiles for devices need to be specific to the device they describe. Monitor profiles are critical to even tell what your images actually look like, and printer profiles make sure that what you see is what you get, but what about scanners and digital cameras?
Let's look at scanners first. Profiling a scanner consists of producing an image of a calibration target that contains patches of known colors and comparing the result to what those colors are supposed to be. The software can then calculate what is needed to correct each color, recording the results as a profile. The most common program used to create profiles for desktop scanners is Monaco (now X-Rite) EZColor. The program comes with a reflective IT-8 target for profiling desktop scanners but you will need to buy an IT-8 transparency target to profile a slide scanner. Depending on what kind of scanner you have, you may be able to tell it to tag the files it produces with this profile or you may have to do it yourself once the file has been created. To do this in Photoshop, use Edit >> Assign Profile. This won't alter the actual image data in the file but will alter the image's appearance by correcting how the color data gets interpreted. After doing this, you should then use Edit >> Convert to Profile to convert the image data back to the profile you use for your working space.
Profiling a scanner can be reasonably effective, but should be considered optional. Even without an input profile for your scanner, so long as you get an image looking how you want it to on your calibrated monitor, you can be assured that the RGB colors in the file truly represent the colors you see. The color management system built into your computer's operating system automatically translates from the color space of your image (your "working space") to the color space for your monitor. This way, no matter where you get the image you start with and no matter how you get it into Photoshop, you can edit it with confidence, knowing that your final file looks the way you think it does. A scanner profile might get you a closer starting point, but once you start editing it you have exactly the same ability to see and control color with or without a input profile.
Profiling a camera is much more problematic though.
White balance is the obvious challenge that needs to be overcome to effectively profile a camera. The process of creating the profile is much as it is with a scanner. A target is photographed and the resulting image compared against what it should look like, with the corrections needed for each color then written to a file to create the profile. But the color of the light illuminating a scene alters the colors that will get recorded when we photograph that scene. As a nature photographer, I want to record the "golden light" of sunrise and sunset but if your aim is to record accurate color, white balance can be a big problem. Only if the lighting is identical to what it was when the target used to create the profile was photographed will the profile be accurate. Indeed, if the more the color temperature differs, the less useful the profile will be for getting accurate color. Studio photographers who shoot under controlled lighting situations may be able to avoid worrying about color temperature, but those of us who work outdoors can't. This alone is reason enough that I have never tried profiling any digital camera I've owned.
But if you do, white balance isn't the only hurdle you will need to overcome. Digital cameras don't actually see in color, the actual raw data being recorded in an odd pattern known as a Bayer Mosaic. Sophisticated software, either in the camera or in your raw converter "de-mosaics" that raw data to form the colors we see in the rendered final image. But digital cameras don't record brightness the way we see it either. The individual photo-sites merely count the number of photons falling on them in a strictly linear fashion. The human eye though is sensitive to brightness in a very non-linear fashion, registering successive doublings of luminance as being equal changes of brightness. The linear gamma of the raw data has to be tone mapped into a brightness curve that looks more natural to us. Taken together, all this means that any attempt to profile a camera is at the mercy of the software used to render it. For a profile to be completely valid, we need to ensure that all camera and raw conversion parameters are the same as they were when the profile was made. As with white balance, this is rarely practical except for those shooting under controlled conditions.
By and large, cameras and converters already try to create results that conform to standard color spaces such as sRGB or Adobe RGB. The actual raw data has no color space since it isn't even yet in RGB format. Some profile must be chosen for the final image since the RGB numbers must mean something. Camera settings typically let the user select this profile, but it doesn't actually get applied until the raw sensor data is rendered as an actual image, whether that is an in-camera jpeg or some other format on your computer. If you now try to assign a camera profile of your own creation, it must be done on top of what the camera or raw converter already does, not in place of it. This means you basically end up fighting against the cameras efforts to render the scene with the standard color space encoding. True, using Hue/Saturation, Curves, Levels and other controls in Photoshop instead can be seen in a similar light, but the use of an input profile is often portrayed as being less destructive which may or may not be true.
Some cameras and raw converters even go one step further in that they automatically adapt how they convert images to the content of the image, subtly re-lighting a scene or biasing various colors in their attempt to create the most pleasing rendering possible. A given color in one part of an image may get rendered differently than the same color elsewhere in the same image, depending on what is next to it. Only by turning all such features off can a profile truly work as intended, but doing so generally will result in a less compelling image, even if in some ways it may be more true to the original subject.
Personally, I don't feel a need to profile my cameras. But there are a number of packages available if you wish to give it a try.