| Home .:. Intro .:. The Color Correction Problem .:. The New Spectrum for Accuracy .:. Applying the Spectrum |
|
The Color Correction Problem What is it about copying an image from one medium (or device) to another that makes it so difficult to get an accurate result? What factors are responsible for the distortions in the copied image which are common to printing, scanning, and film and digital photography? There are many complex issues that contribute to this problem. But these issues can largely be reduced to just a couple of problems central to converting images between mediums. The first big issue is this: can you get the machine your working with to display the colors you tell it to? In other words, if you have a color in mind (represented by numbers perhaps), do you have the means to get the device to show this color? Any device will have a system for representing colors. Monitors will use RGB (measures of red, green, and blue light) to represent each color. But RGB values that represent one color on one screen, will not necessarily represent the same color on another screen. Thus, a color profile for each device must be created, that is, a large index of RGB values for each machine that shows how colors may be matched between them. Creating color profiles is a lengthy and laborious task. While this process can certainly be facilitated by Master Colors' ideas, it is still not even the most significant contribution it offers. The true power of these ideas lies in their ability to resolve the second big problem of color correction. Once color profiles for all the devices in question are established, and we now have total control over displaying the colors available, there is another stumbling block which can prove even greater than the first. Are we able to show on one device all of the colors available to the other? In other words, some colors in our image, residing on one device, may contain colors that simply cannot be displayed by the other device, onto which we would like to copy the image. This is because the range of colors available to each device, the gamuts, will often differ, sometimes quite significantly. For example, the range of color available through paper and ink (as with a printer) is quite different from the range available on a monitor. You may not have noticed, but generally the colors on a printout will be much darker than those on the screen. The monitor is its own light source, and provides a consistently bright range. The printout is at the mercy of an external light source, so unless view under a high-wattage bulb, or bright daylight, the colors you will see are going to be considerably darker. Thus, an inherent limitation of color correction is revealed. When you print out a bright color from the screen onto paper, the printed color and the screen color are not the "same" color. The color on the page is simply an approximation of the screen color, chosen from the printer's built-in range. This process of color approximation is what causes the great complexities of color correction. You may not have noticed that the printed image was darker, but what you probably have noticed at some point is that the image seems funny looking, a little distorted, or not what you had in mind. These notorious distortions are a byproduct of the printer's approximation of the colors displayed on the monitor. Let's examine why these distortions occur. When we copy an image to a device with a significantly different range of available colors, we need to make sure each color in the image now fits in this range. In other words, the colors must be changed in order to be displayed on this device. Let's look at how this is generally accomplished. The industry standard for describing any color in any medium is through the CIE L*a*b* color space (L*a*b* for short). Using L*a*b* coordinates, we are able to represent any color in the visible light spectrum. So technically, there is a universal L*a*b* space which contains every conceivable color. A color can be represented somewhere in this space, regardless of which device it appears on.  An approximation of the shape of the L*a*b* color space. Lightness (L) is the vertical element, while the factors of hue and intensity (a* and b*) are horizontal. Note that this is a three-dimensional space. Naturally, any device will only be capable of showing a fraction of the colors in this universal space, dictated by the gamut of the device. So, the color space that the device uses will resemble a small pocket within this greater space. A device which supports lighter colors will occupy a region higher up in this space, while one with darker colors will have a lower region. |
Left: Region of the greater L*a*b* space occupied by the device with the lighter color range. Only the colors confined in the enclosure are accessible by the device. Right: similarly, the sub-space of colors available to a device with a darker range.
|
When copying an image that lies in one of these color ranges over to the other color range, the basic process is to pick colors in the new range which are locally of the same coordinates of the original colors. For example, if moving the image to a darker medium, the colors are basically just "pushed down" to fit in the darker range.  An illustration of the process of bringing an image into a new color range. First, all the colors in the original image are located in the color range sub-space. Then, all of those locations are shifted down to new L*a*b* positions, fitting in the darker range. The colors indicated by these new positions comprise the palette of the copied image, thus making the image darker, and able to be presented on this device. This depiction is a bit of an over-simplification, but it is the basic approach to copying an image to a new, differing medium. It seems quite straightforward, and you may wonder why this approach would lead to problems. At a glance, does appear to be the most logical solution to the given problem. But there are hidden complexities involved. Part of the problem lies in the inherent limitations of the color space involved, the L*a*b* space. It is a great system for objectively cataloging and identifying colors, independent of the device. However, its strength in these areas has largely to do with the fact that its language of color description is one that is convenient for machines, for the purpose of measurement. It is not a space which a human would recognize as a visually intuitive means of organizing and describing color. So, it should not be surprising that the results of color transformations and adjustments performed within this space do not always agree with human judgment. For instance, by merely "pushing down" a color through this space (reducing its 'L' attribute), this does not guarantee that the other characteristics of this color, valued by human judgment, will be conserved. The a* and b* coordinates will not change, but these are not necessarily the attributes of color we wish to see remain constant during this shift. In all likelihood, we will see the color become slightly distorted in ways we don't expect or want. This is part of the reason for distortions we see. The solution to this is to make our color adjustments in a different kind of color space. One that does not distort colors in unexpected ways as we adjust their coordinates. In essence, any adjustment of colors through such a space will yield results which are in precise accord with our intuitive expectations. Such a space is known as a perceptual color space. (a more thorough description of a perceptual color space is found here) We will discuss in more detail how a perceptual color space aids the color correction process in the next section. In addition to the natural distortions involved with the L*a*b* space, or other spaces more inclined towards the technical measurement of color, there is another fundamental problem with the image's transition. In the above example, when we brought the image down into the darker color range, we assumed this range was the exact same size as the previous range. Consequently, all of our colors were able to fit nicely into the boundaries of the new range. In practice, cases are seldom this convenient. Quite often, the new range will be compressed, off-line, or misshapen, causing some of the colors to "hang off" of the edges. |
| The colors which exceed the boundaries of these contracted spaces present a dilemma. How can we get the device to display the overflowing colors, and still make them appear as similar as they did in the original image? The most crude and straightforward solution is just to push all the overflowing colors into the nearest border of the space. |
|
The results of this approach should be fairly obvious. It will certainly result in a distortion of the image's colors. The further the colors must be pushed, the greater the distortion. Consequently, the more extreme colors in the image are dulled, as the brightest, lightest, or darkest colors must be pushed into the new range. This will definitely obscure much of the nuance of the image. This is an extreme solution to the problem, and it's unlikely that any existing correction process employs a strategy so simple. But it does illustrate that some distortion will be inevitable in these situations, and that a compromise is needed between this approach, and others of more sensitivity. What Master Color's has provided is a complete clarity of terms under which such a compromise is reached. A precise balance can be struck between the need to fit all colors in the new space, and the minimization of the accompanying distortions among the colors. Such a powerful compromise is an outgrowth of employing a perceptual color space, and will ultimately serve as the dictum for what accuracy truly is. We will discuss the nature of Master Colors' solutions to these central problems of color correction in the following section. Next section: The New Spectrum for Accuracy |
