Understanding How Your Camera and Computer Handle Color

"The Woods Behind My House" - Unedited photograph, Quaker Oats Box Pinhole Camera (6th Grade Art Project), 60 Second Exposure, 8x10 Negative, 8x10 Contact Print, Processing done by hand in the boiler room
"The Woods Behind My House" - Unedited photograph, Quaker Oats Box Pinhole Camera (6th Grade Art Project), 60 Second Exposure, 8x10 Negative, 8x10 Contact Print, Processing done by hand in the boiler room | Source

Two Roads

You know that line from the famous Robert Frost poem? "Two roads diverged in a wood..." Well, there are two roads you can take with your colors once they're in digital form. It's important that you understand the difference so you can knowingly choose which road to travel.

A digital pixel uses numbers to quantify color. It's usually three numbers, one each for Red, Green and Blue. For most photographic computer image files, each of the three numbers range from 0 to 255. But what do these numbers really mean? The answer might surprise you. For "Untagged" image files (i.e. those without an ICC profile), those numbers actually have NO WELL DEFINED MEANING! This is the road of "Unmanaged" color. It's fraught with confusion and frustration. On this unmanaged road, the color that's associated with a given set of RGB values can change between different computers, monitors, printers and projectors. There is no consistency. Each device is deciding on its own what that color should be. It's mayhem! And yet this road is still very common. It's hard to believe, but even Art Shows and Contests that use digital images for their initial judging are for the most part still on this unmanaged road.

The alternative is the "Color Managed" road. Going down this road can seem daunting and confusing at first, but the rewards are significant. Using color management, a set of RGB values DOES have well defined meaning. Having an ICC profile embedded in an image file explicitly defines what the RGB values in that file truly mean. This allows you to communicate color accurately between different computers, monitors, printers and projectors. Using color management, you can share an image file across the world with the expectation that your colors can be handled accurately on the other side. But before embarking on the road to color management, it can be helpful to find solid footing on some underlying fundamentals.

Rainbows and Prisms

The colors that humans can see are made up of electromagnetic waves ranging in wavelength from 380nm to 730nm. This is called the visible spectrum. This range varies somewhat for different people, but not by much. For this first concept, we are going to use this range of wavelengths but otherwise ignore human perception and just measure color from a physics viewpoint. We analyze the physics of a color by splitting its light (like a rainbow or prism does) and then measuring the electromagnetic power at each wavelength in the visible spectrum. The resulting spectral power distribution data yields a very precise definition of that color at a physics level. There are actually an infinite number of different wavelengths in the visible spectrum, but color measuring devices (called spectrophotometers) combine groups of wavelengths into some number of discrete "buckets." Some devices separate the spectrum into 36 buckets (one bucket for every 10nm in the visible spectrum), some 72, some even more than that. This kind of color measurement in the physics domain is known as "radiometrics" and DOES NOT involve any analysis of human perception. The spectral power distribution data accurately describes the physics of a color, but how will an average human actually perceive this color?

A Sprectrophotometer's Measurement Data for a Specific Color

What Happened in 1931

Leading up to 1931, several independent teams of researchers tested human color vision using groups of individuals. Inside the retina of the human eye there are three unique color receptors that generate electrical impulses when stimulated by different (but overlapping) ranges of wavelengths, roughly corresponding to what we perceive as Red, Green, and Blue. The researchers wanted mathematical functions to describe the three response curves. They would display a color that was a known wavelength and have each person turn three dials to adjust the power levels of a mixture of known wavelengths of Red, Green & Blue projection lights until the person perceived a match to the control color. They did this over and over again with a range of colors. The research teams then used their mass of data to build mathematical human spectral sensitivity functions. These functions can be used to take the spectral power distribution data for a specific color and map it into a 3d mathematical space that quantifies the color that an average human observer will perceive. The human response functions from the independent teams matched well and the combined result was the definition of the CIE 1931 "Standard Observer" functions and the CIE 1931 "XYZ" three dimensional color space. Analysis of color in this domain is known as "photometrics" and is all about human perception. Further research created other useful mathematical 3d spaces from this same data, like the CIE 1976 Lab space. This space is a transformation of the XYZ data that makes the distances between numbers be "perceptually uniform" which allows us to more easily quantify the perceptual difference between colors.

Here's a key point to understand between radiometrics and photometrics: different sets of spectral power distributions can actually map to the same human perceived color. That's right, there are lots of different spectral distribution mixes that can make a human perceive the same color! They are called "Metamers." So, knowing the XYZ or Lab values for a color will let you know how it will be perceived by an average human observer, but they won't let you map that color back to a specific set of spectral power distribution values that originally generated the color because there are a lot of them, and there's no way to figure out which one generated the color in the first place. However, by knowing just those three numbers in either the XYZ or Lab spaces, different color reproduction devices can use their own measured spectral power distribution generating capabilities to create the same "perceived" color. Or, if they don't have the capability to hit that exact color, they can at least intelligently do their best to get as close as possible.

Communicating Our Colors

In 1994, an international standard was developed to allow different computer systems to convey color accurately to each other. This was the birth of the ICC profile. The profile gets "embedded" inside an image file. It contains a set of equations and lookup tables for translating between the pixel data in the file (usually RGB) and standard XYZ or Lab values that represent human perception. After that translation is performed, then any other device will know how the file's colors are meant to be perceived and those devices can do their best to generate the correct colors. There's a lot of complexity summed up in the simple statement, they'll "do their best" to generate the correct colors. What is white? What is black? What should I do if I'm not able to generate an exact match? Should I shift all the colors a little bit to keep them coordinated, or just clip those out of gamut colors off abruptly? Am I trying to be absolutely colorimetrically accurate, or is being perceptually pleasing more important? etc. etc. The complexity is daunting, but hopefully less so if you understand some of these fundamentals.

The first practical steps you should take on the road to color management are the following: Set your image editing software to make use of ICC profiles, set your default working color space to a standard ICC profile such as sRGB, AdobeRGB, or ProPhotoRGB, and get your display calibrated to a good state and profiled using a color measuring device with associated software that creates an ICC profile specific to your display. The next step is to use ICC profiles specific to your various combinations of printer settings and papers.

What About My Camera

Digital camera sensors are not spectrophotometers. They don't split light into 36 or more discrete buckets of wavelengths for accurately defining color at the physics level. They have three overlapping RGB sensitivity ranges similar to, but NOT identical to, the human retina. So, each different camera sensor has it's own unique set of "observer" functions. We help the sensor get closer to human perception by setting its "White Balance" to match the light source in the scene, like tungsten or cloudy daylight or flash. Or we tell it explicitly what white is supposed to be by setting a manual white balance, or we just let the camera make its best automated guess. The important thing to understand is this: digital cameras DO NOT produce "accurate" color by default. All the various camera processing options, picture "styles," RAW workflows, etc. are aimed at creating "pleasing" colors for the type of scene that's being photographed. Even with identical lenses, identical white balance and identical exposure settings, different camera models will produce a different set of colors for exactly the same shot.

So you ask, if the colors from the camera aren't accurate in the first place, why use color management at all? Good question. Here's the answer: As the image comes out of the camera, it's best to put it directly into a well defined ICC profile color space (like sRGB, AdobeRGB, or ProPhotoRGB) so that any further editing you perform will have well defined results. Also, being in a well defined ICC profile color space makes it so the image you see on your color managed monitor is well defined, can be recreated accurately on a print, and can be recreated accurately on someone else's monitor, projector, or printer (as long as those are also color managed).

Also, for very specific situations (like art reproduction), you CAN use color management to get a camera to produce colorimetrically accurate results. But this takes a fixed lighting setup, fixed camera settings, and a profile of the camera's color space built using targets that have been measured with a spectrophotometer. Profiling software uses the camera's shot of the targets together with the measured data of those targets to build an accurate mapping from the camera's data values into standard XYZ or Lab spaces. From there they can be accurately translated into standard ICC profile spaces and viewed or printed accurately on color managed systems.

Example 2584 Perceptually Uniform Patch Set to be Measured with a Spectrophotometer and used for Profiling a Printer and a Camera
Example 2584 Perceptually Uniform Patch Set to be Measured with a Spectrophotometer and used for Profiling a Printer and a Camera | Source

What Do You See?

Here are some images of an oil painting and a mixed media piece that have been shot with a profiled camera setup and translated accurately into the sRGB color space. If you have a color managed system with a color managed image viewing application and a color managed display, you should see a pretty accurate rendition of these pieces. Without them, I have no idea what colors you are actually seeing!

"Passing the Old Homestead" - 16"x24" Oil on Canvas by Roy Grinnell
"Passing the Old Homestead" - 16"x24" Oil on Canvas by Roy Grinnell | Source
"Ruffled" - 18"x26" Mixed Media on Canvas by Jeanette Korab
"Ruffled" - 18"x26" Mixed Media on Canvas by Jeanette Korab | Source
Detail of "Ruffled" - 18"x26" Mixed Media on Canvas by Jeanette Korab
Detail of "Ruffled" - 18"x26" Mixed Media on Canvas by Jeanette Korab | Source

Here's an image from my family room that doesn't have "accurate" color, but has been processed by a RAW workflow to have what I consider "pleasing" color in the sRGB color space using a color managed system. If you too have a color managed system, you'll see the colors as intended. Otherwise, I have no idea what colors you are actually seeing!

"Diane saying, 'Yes, I won this BRAND NEW TV in a contest!!'"
"Diane saying, 'Yes, I won this BRAND NEW TV in a contest!!'" | Source

Why sRGB for these Examples

The sRGB color space has a fairly small gamut compared to AdobeRGB and ProPhotoRGB. It's not normally used for art reproduction because some vibrant art colors will be outside its gamut. Also, today's printers and displays can generally reproduce colors that are outside its gamut, so using sRGB is too restrictive. However, sRGB is an old, well established color space that serves somewhat as a "common denominator" for a lot of systems, even ones that are not color managed. So, when sharing image files with unknown destinations, I translate them into the sRGB color space because it's still the safest choice. By putting these files in sRGB, I know you have the best chance of seeing them as intended even if you don't have a color managed system. But without color management on your system, I still have no real idea what colors you're seeing.

Your Choice

Hopefully, the historical and technical background provided in this article can help you find your footing as you choose which road to travel when it comes to digital color.

Color Management Quiz


Which concept helped you gain new insight into color management?

  • Description of "unmanaged" color
  • Description of how a spectrophotometer measures color
  • Description of the 1931 human vision research
  • Description of how an ICC profile works
  • Description of getting accurate colors from a camera
See results without voting

More by this Author

  • How to Understand Depth of Field

    Many Depth of Field explanations leave people confused in one way or another. This article offers a unique approach to help you understand WHY Depth of Field behaves the way it does.


No comments yet.

    Sign in or sign up and post using a HubPages Network account.

    0 of 8192 characters used
    Post Comment

    No HTML is allowed in comments, but URLs will be hyperlinked. Comments are not for promoting your articles or other sites.

    Click to Rate This Article