Complete Color Integrity
We begin by explaining color integrity in simple terms. Following that there are several routes to take toward a more thorough understanding of color integrity, and the reader can follow whichever routes are familiar and comfortable. I need to be clear from the start, however, that I have nothing against the creative, artistic use of color in photography. However, I do believe that it is important to start with an image that has color integrity before getting creative with the color. Otherwise a lot of creative energy is wasted in trying to compensate for what the image lacks.
The concept of color integrity deals with the behavior of color as it appears when in bright areas of a scene and when in darker areas of the same scene. Imagine a colored patch placed in a dark area of a scene. If I move this patch to a light area of the same scene the patch becomes lighter but it does not change its apparent color. (Of course this model assumes the light source itself is not different in the light and dark areas of the scene, but cases where the light does change are easily built upon this framework.) If we place this colored patch in an area where it is shaded at one side and the light is unveiled gradually to full brightness the other side, again the color of the gradient does not change as it becomes lighter. This is how the eye expects colors to behave, whether looking at the folds of a drape, the leaves of a tree, or the contours of a face. When the shading of colors in an image behaves in this natural manner we call that color integrity. When the colors in an image change unnaturally going from dark to light the eye becomes confused and the lack of color integrity can produce an unsettling effect.
In photography there are two distinct steps involved in producing images with this color integrity. First we must produce a digital image that corresponds to the colors and variations of color in the original scene with acceptable accuracy. This step of creating the initial accurate image having color integrity is called calibration. We find it best to use a calibration system based on a stepped grayscale. In digital photography gray is composed of equal parts of Red, Green, and Blue so calibrating with a stepped grayscale exercises all three color channels identically and because the grayscale steps through the range from black to white the calibration covers the entire response of the digital camera sensors (or the film) over the complete range of all three colors, red, green, and blue, that the digital camera or film sees. Using a correct grayscale calibration will produce an image file that starts out with color integrity. We have a web page on grayscale calibration and the ColorNeg and ColorPos plug-ins include several methods of grayscale calibration. For those who are curious about the choice of grayscale calibration and whether it is really adequate and preferable to using colored patches, in addition to the web links, Comments About Calibrating Digital Images in this document has more detail on calibration and why many of the current methods of calibrating digital cameras and film, particularly "profiling," often lose color integrity right from the start.
The second step in producing images with color integrity is in making sure that adjustments, especially initial major adjustments, to this initial calibrated image can be made while maintaining its color integrity. Here we are particularly interested in adjustments which lighten dark areas of the image or darken light areas of the image and also in adjustments to produce color balance, both general and in localized areas of the image. This is where my new discoveries come into play.
To explore how we do this, let us take a calibrated image fresh from a digital camera in which the colors and their shadings do behave as described above that is, a starting image that has color integrity. We want to make the image darker, but we want to maintain color integrity as we do so. Above we moved an imaginary colored patch into different lighting conditions to see how it behaved. Now it is helpful to think of each tiny bit of color in our image is such a "patch." Whatever happened to the color of our reference patch when we made it darker by moving it into areas with dimmer lighting, that same thing has to happen uniformly to all the colored "patches" in our image. As we prove elsewhere on these pages, doing this is very simple we just add black. To darken the "patches" of color that make up the image we must add black to each color in the same way that an artist-painter might add black paint to darken a colored paint on the palette. To darken the entire image uniformly we need to mix in the same amount of black to the color of each color patch in the image, resulting in exactly the same effect as if we had dimmed the light source. This idea of "adding black" to darken an image probably seems obvious now that it has been explained and is not at all difficult to do to a digital image. But you won't find anything similar to "add black" in the Photoshop plethora of commands nor is it particularly easy to ferret out exactly how you might achieve that result in Photoshop. Even then the tool may work in some cases and not in others. Apparently this "obvious" approach really is not all that obvious!
Suppose now, instead of darkening the image, we wish to make it lighter. Again, it is very easy to achieve that. We simply reverse the above and remove black instead of adding it. To uniformly lighten the image we remove the same amount of "black paint" from the color of each patch of color in the image. Again, this is easy to do for a digital image, just the reverse of adding black. A painter would find it difficult to remove black paint from an already mixed color!
"Adding black" also plays a less obvious part in properly dealing with color. The terms "white balance" and "color balance" are commonly used in dealing with the fact that the human eye automatically "accommodates" so that it sees as white or nearly so whatever light source predominates in any scene under view. Neither photographic film nor the sensors in digital cameras do this and so the image nearly always must be adjusted white balanced or color balanced to achieve the same effect as the eye's accommodation. To understand what is happening here we need to look at the scene in the way the camera does, with three image sensors, one for Red light, one for Green light, and one for Blue light. When this set of sensors produces an image under what it regards as "white" light the resulting calibrated and uncorrected digital image will look correct when properly displayed. However, suppose the light is really reddish (like a regular light bulb, compared to daylight). The human eye will compensate, so the scene still will look normal, but the resulting digital image will appear reddish because the Red component of the digital image is too bright. We can deal with this just as we do when the entire scene is too bright. We found that adding black to the color of a color patch produces the same effect that dimming of the original light source would have done. In this case, however, only the red component of the patch is too bright, so we add black only to the red component rather than to all the colors of the color patch. In this way we dim the red only and exactly compensate for the non-whiteness of the light source in the scene. Of course we can use the same procedure of adding or removing different amounts of black to the red component, to the green component and to the blue component in order to color balance the image with a result that is similar to what the eye sees automatically. Again what we describe here is something an artist painter cannot do, but in this case it is no great loss since the artist finds it unnecessary. The eyes serve as sensors for a painter and so the artist has no problem similar to photographic color balance in the first place.
In the above we have learned that by using actions which effectively "add black" or "remove black" from the colored "patches" (pixels) in a digital image we preserve and maintain the color integrity of an image. Black is colorless, so this makes sense and in fact seems obvious. But white is also colorless. Is "adding white" the same as "removing black"? The answer is no. When we say "adding" here we do not mean "adding" in the mathematical sense, but in the sense of taking some colored paint and adding white paint to it or adding black paint to it. If you add some black paint to some colored paint and find that you have added too much black, adding white paint will not reverse the effect and get you back where you wanted to be. Seeing this in your mind is easier if you take an extreme case. Suppose you had a spoonful of red paint and you mixed in into a cup full of black paint, so the result was nearly black with a very slight red tinge to it. If you mixed a cup of white paint into that, clearly you would not get back your original red color, but instead a middle gray (black plus white) with only a slight red tint to it.
Nonetheless, as you would expect, white has no color and so "adding white" is another tool that preserves color integrity. The reason that "adding black" works is that it makes color "patches" behave as they would as the image lighting produces different levels of dark and lightness in the scene and so "adding black" is recognized by the eye as being natural. "Adding white" turns out to be exactly what happens in the areas of a scene that are bright enough to produce glare (or specular reflection). As the glare increases in bright objects more and more white is added to the color in the color patches. "Adding white" also turns out to be exactly what happens in a natural scene where there is fog. The individual "color patches" in the scene are the color that they would be without the fog but with white added to them. Again, just as it is possible to "remove black" in digital images it also is quite possible to "remove white" and indeed as is shown in the Fog Example section of this document, removing white from an image of a foggy scene can do a quite convincing job of removing the fog. You will not find any tool similar to "add white" in Photoshop. Unlike black, the tool for adding white is not just hidden in an unexpected place, it simply is not there. You can sometimes combine several actions and achieve the action of adding white but even that is not dependable.
In summary we have learned that to have color integrity in a final image the first step is to get the image into the computer using a calibrated device, and that the most problem-free calibration is done using a grayscale. The resulting image will have color integrity as a starting point. Then the major adjustments of color balance and of the tonal scale should be done using tools which preserve the color integrity. We have found that the required tools are those which can add and remove black and those which can add and remove white to the color in the "color patches" that make up the image. For tonal scale adjustments there is nothing saying that there must be a consistent amount of white or black added to or removed from each of the color patches, so adjustment of the tonal scale can be made quite differently in different parts of the image without affecting color integrity but only if it is done by adding or removing black or white to or from the color that would be in the patch. In the case of adding black in different amounts to the single color channels of Red, Green, and Blue in order to achieve color balance, there does need to be a consistent amount of black added to groups of patches which have the same light source.
So, the starting point for color integrity is a properly calibrated image which accurately represents the image colors as they appear in darker and lighter areas. Then to maintain color integrity as you adjust the image or parts of the image, the rule is to work by adding or removing black or white. Once you have grasped it, this seems such an obvious rule that surely it must be a core principle behind digital photography. Yet as this is being written, far from being a core principle, it is so completely unknown that in Photoshop, acknowledged as the premier image editing program, the tools to perform these four basic actions range from difficult to find to nearly impossible to achieve. In general, other image editing programs fare no better. In addition, many actions taken under the guise of "Color Management" actually subvert the well-planned ICC design. As one result, it has become more and more difficult to get an image from a digital camera into a computer without having color integrity lost from the start. Our ColorNeg and ColorPos Photoshop plug-ins are capable of adding and removing black and white, but there is room for improvement as a result of this study.
Perspectives and Comments
Genesis of the Idea what led to the discovery of these simple facts.
Why We Give Few Illustrative Examples of Color Integrity
Color Integrity from the Viewpoint of Artistic Painting.
Color Integrity from the Viewpoint of Basic Physics and Mathematics.
Trying to Deal With Color Integrity in Photoshop
Color Integrity and Color Balance A Few Examples
Comments on Calibrating Digital Images
"Acceptable Accuracy" in Calibration
Calibration and Color Profiling
The Pitfalls of Using Profiling as Camera Calibration
Much of the material in this document and the analyses are original with us. This document is
Copyright © 2009 by C F Systems
If you plan to use our original material or analyses in any published form, please contact us at www.c-f-systems.com for terms and conditions. Before-the-fact terms typically require no more than appropriate acknowledgement.
Go to C F Systems Home Page