Color Balance, White Balance, Calibration, and Color Temperature
There is confusion in the use of the terms "color balance," "white balance," and "calibration" even among experts. These terms are sometimes used interchangeably and, worse, very often they are combined together as a single concept. They are different and understanding the difference makes it much easier to successfully and consistently deal with digital color photography. In this discussion, all of these terms involve the treatment of three primary color systems red, green, blue (RGB) or cyan, magenta, yellow (CMY) that are central to producing film and digital images. It is important to separate these three primary systems from other multi-color systems such as cyan, magenta, yellow, black (CMYK) used in printing and the YIQ used in video, etc., where the same terms may be used in quite different ways.
Color Balance and White Balance
White balance and color balance are closely related, but they are not the same thing; color balance is the more general term and white balance is essentially one method of balancing color. The concept of color balance exists at all only because the human eye automatically accommodates to the color of the dominant light source so that familiar objects tend to look about the same color whether illuminated by sunlight, shade, indoor lighting, etc. Neither film nor the basic sensors of digital cameras do this, so for the colors in an image to look correct, a compensation must be made. Professional film photographers and (wet) darkroom workers do this using color compensation filters (cc) in various densities of the primary colors. This is exactly the right thing to do because using filters is precisely equivalent to adjusting the color of the light source itself rather than adjusting the color of individual objects in a scene. Putting a color filter over the lens of a camera is exactly the same as putting the same filter over the light source (if that is physically possible). Filters effectively adjust the sensitivity of the film or print paper to a primary (red, green, or blue) by the same proportion (percentage) whether the light is dim or bright, so the action of a filter can be easily calculated by a computer or digital camera (see our page on Color Integrity for a detailed explanation of this). The right filter can precisely correct for a light source that is too red, too blue, too yellow, etc. whether a real filter is used or a percentage adjustment calculation is made on each pixel in an image.
Because an image is formed from the light reflected by objects in a scene, the color of the source lighting is normally what affects color balance. The most obvious way of achieving color balance is to measure the color of the light source and adjust the image accordingly. Because the eye accommodates, the lighting for any scene should be "white," so the filter or filter-equivalent adjustment of an image to correct the measured light source to a standard "white" is called "white balance." This is white balance in its original, true sense. A device is arranged to capture light from the light source and measurements typically are made through a suitable diffusion disk. The camera itself can be used as the measuring device by putting a diffusion disk over the lens, pointing the camera in the general direction of the light source and taking a photograph. This is in fact an excellent measurement of the color of the source lighting and normally will result in a very good color balance. In situations where it is practical, it remains the best method of achieving a good, standard, color balance.
This true white balance has been subverted and now many other color balance methods are also called "white balance" even though they are not direct measurements of the color of the source lighting. In digital cameras, "white balance" is usually measured from the scene itself; that is, from reflected light rather than directly from the light source. The reflected light is typically measured according to some proprietary averaging algorithm over the image pixels, biased heavily toward some selection of the lighter pixels that appear not to be highly colored. How well this works depends upon how average the scene is, particularly in the highlights. In Photoshop, the Levels tool has a "white picker" that is often used for a "white balance" by clicking it the whitest area in the image. While this method sounds like it should work, more often it does not and it is typically followed by a "middle gray balance" or "middle gray adjustment" to get the colors "right." In our page on Color Integrity we explain that what this often does is throw the color balance off in a way that is subsequently very difficult to correct.
For most calibrated images (see below) color balance is most easily achieved in Photoshop using the Levels tool white picker in an entirely different manner, as explained in our item on Color Balancing a calibrated image. At this point the color balance can easily be tweaked to produce what are usually called "warm tones" by adjusting the red highlight slider on the Levels tool or "cool tones" by adjusting the blue slider. Warm and cool tones represent what is often perceived when a scene is first presented and before the eye has fully accommodated. These effects and normal color balance are easily achieved when working with a properly calibrated image.
Calibration
Calibration and color balance are distinctly different, independent concepts. Color balance has entirely to do with adjusting for the dominant lighting of a scene and separate, different color balance adjustments may easily be required for every image - certainly for every image taken under different lighting. Calibration is entirely to do with the sensing devices used to take the image; camera, film, digital sensors in digital cameras. The calibration stays the same as long as the sensing equipment stays the same and so in general the calibration may be the same for many images. For instance, calibration is normally consistent for the same film, given consistent processing (that is, different rolls of the same speed of the same type of the same brand of film will calibrate nearly identically). Calibration insures that the intensities represented in the image for each primary color match the corresponding intensities reflected from objects in the original scene. Intensities are said to match if for each primary color they are consistently in proportion between the scene and image. An image which is in calibration is said to have color integrity, although it may not be in proper color balance. Images which have color integrity can be easily and accurately color balanced (see for example Color Balancing a calibrated image) and images which do not have color integrity are difficult to color balance at best.
It Is Important to Separate Calibration and Color Balance
Starting from scratch with new (or different) equipment, images will generally be both out of calibration and not in color balance. To get a good result it is necessary to adjust for both. One mistake very often made by amateurs and professionals alike is do both adjustments using the same tools and at the same time. Since these are independent phenomena that arise from completely different causes, if you do an overall, simultaneous adjustment, the adjustments will interact and the calibration adjustments will try to compensate for imperfections in the color balance and the color balance adjustments will try to correct problems in the calibration. The result may be a more exact match to the test image, but that will not translate to other images. There will be poor color integrity in many images based on this calibration or there will be the necessity to perpetually recalibrate. Poor color integrity - poor calibration - shows as difficulty in color balancing future images after a calibration has been made. The most common symptoms are off-color or tinted highlights when the rest of the image looks good, muddy colors in the midtones or shadows when the highlights look good, or the requirement to do a Levels ''middle gray" adjustment to get colors that seem more acceptable.
It is not difficult to keep calibration and color balance separate. Calibration is used to set the tonal light-to-dark range correctly. Consider a red target which has steps which run from nearly black to very intense red, each one having successively more red. Whether done with gamma and blackpoint or with a more complicated Tonal Reproduction Curve (TRC), the objective is the same: to make the red in each step in the image match (as explained above) the red in each step in the target. The same is required for green and for blue. In general the red, green and blue calibrations are independent of one another (although for most digital cameras they are not independent). Even so, it is completely satisfactory and usually preferable to do the red, green, and blue calibrations at the same time. When combined together in the right proportions, the red, green, and blue form gray, so the combined targets form the familiar stepped grayscale. After the component red, green, and blue steps of the gray scale have been calibrated so that each matches (proportionally) the target, color balance becomes simply a matter of adjusting one of the steps by multiplying the red in that step by r and the blue in that step by b so that the step is gray (that is, so that the red, green, and blue are colorimetrically equal). Then, when all the other red steps are multiplied by the same r and all the other blue steps are multiplied by the same b, and the entire scale becomes gray - if the calibration was done correctly.
Does this mean that the entire calibration can be done using just a stepped grayscale? Yes. Moreover, if the calibration is not being done using a stepped grayscale, it is quite likely that it is not a good calibration. This is true for all equipment using systems of three primary colors: red, green, blue or cyan, magenta, yellow. It is true for all film cameras and almost all digital cameras and scanners. It is not true for printers and video systems which use different multi-color systems. It is possible to calibrate three primary color sensors using colored targets which have all the colors within gamut of the target primaries but such calibrations will be no better than a stepped grayscale calibration and can be considerably worse, as such methods often combine calibration and color balance. Mathematically, the reasons for this are quite straightforward and have to do with independence and degrees of freedom in the three primary system. The gist of this has to do with the fact that a white or light gray spot has every bit as much red in it as a very intense red spot in the image might have. As a simple example, they both might have a red value of 240. Now, if you try to adjust the calibration so the 240 red spot is moved up to be more brilliant, say 250, there is no consistent way of doing this without moving all the other pixels with 240 red to 250 red. This will automatically move the red in the light gray from 240 to 250, giving the white highlights a decidedly pink tinge.
But if I do not use a colored target for calibration, won't some colors be off? That is actually a trick question. The answer is yes, some colors in images produced by the calibrated system will be "off." Some colors will be "off" no matter how the system is calibrated. It is generally mathematically impossible to perfectly match all the colors in a scene (or a test target) by using a three primary system, although the differences may be and normally are small. It is more a matter of controlling how the colors are off. Some colors will be "out of gamut" for the three primary system being used but even more commonly, other colors will be off because the equipment is not sensitive to the visible spectrum in a way that is completely equivalent to your eye. As may not be immediately obvious, there can be some spectrum sensitivity mismatch between the camera and your eye even for the gray in a grayscale. However, if the grayscale target is a good one, correctly made, the mismatch will be the same for all the steps and so will calibrate correctly. Dunthorn Calibration describes methods of calibrating various equipment using a grayscale.
For artistic reasons it is quite possible that you really may want to have the colors more vibrant - or more pastel - or the reds more pure - etc. than your calibrated images routinely produce. Neither calibration nor color balance is the place to do that. Once the colors are as correct as your equipment can make them, explore the several ways of making them more vibrant - or more pastel - either overall or individually for parts of an image.
Color Temperature
Color temperature as used in photography is an outdated concept that once served a purpose and now serves primarily to confuse. The color of the light source is important in photography and in the study of color in general. Back in the early days of color science the physics of "black-body radiation" had recently been given a precise mathematical form so that the spectrum could be readily calculated for light emitted by a black object (i. e., one that doesn't reflect light) raised to a particular temperature. This applies fairly closely to a red hot iron bar, for example (around 1000 – 1200K). Color scientists considered sunlight to behave as a black body around 6000 – 7000K, somewhat higher than estimates of the surface temperature of the sun. This was done even though the light is substantially filtered coming through the atmosphere (consider sunset effects, where the atmospheric filtration is enhanced because the light takes a longer path through the atmosphere). The actual correspondence of sunlight or daylight to black body radiation is far from exact and depends on the circumstances of the day. However, the other common light sources of the era: gas lights and other flames and particularly the various versions of incandescent light bulbs, were good approximations of black bodies. So, the color temperature scale of the color of illumination sources came into being. In this pre-computer era this was a good thing. For example, a floodlamp manufacturer could target a design temperature for the light bulbs as 3200K and at the same time a film manufacturer could design a color film to be properly balanced when illuminated by light with a spectrum calculated for a 3200K black body. Flashbulb and film manufacturers could both target 3200K - or other agree-upon color temperatures - in the very different ways required for their manufacture.
But all light sources are not black bodies and many have entirely different characteristics. Fluorescent lights, electronic flash, and many more recent light sources have very uneven spectra and do not have a color temperature in any meaningful sense of the term. Light which is selectively filtered in any way usually will not have a spectrum that corresponds to any black body temperature. Light reflected from a wall, greenery, etc. will not have a blackbody spectrum. Nonetheless, such sources routinely have been assigned correlated color temperatures, in more recent times usually based upon the a visual red/blue match to a source of known color temperature. Truly odd is the practice - sometimes to the level of a fetish - of associating color temperatures with the bluish and reddish effects occurring in natural outdoor lighting. The cause of these effects is almost entirely atmospheric filtering and selective reflection of sunlight. It has nothing whatsoever to do with blackbody radiation and does not follow the blackbody color temperature curve. In fact, it is more accurate the describe natural lighting from filtered and reflected sunlight as a wide band or perhaps an ellipse in the CIE diagram. The following diagram shows how the color temperature scale tracks through the familiar CIE horseshoe diagram:
As the diagram shows, the blackbody curve is precisely that: a curve. It is not just shifting from red to blue. The amount of green changes by a lesser amount, but it does change all along the curve. We do not show it in this schematic diagram, but on the web are precise plots of the blackbody curve with the CIE Standard Illuminants A, B, etc. plotted along with it. Even these basically incandescent bulb references do not fall on the blackbody curve. Color temperature is a very precise thing as far as the physics of blackbody radiation is concerned, but it is a very imprecise and clumsy tool in digital photography. In fact, its presence in digital photography is very much like the QWERTY keyboard in computing. The reason for the clumsy QWERTY keyboard disappeared within years after it was invented in the 1800s, but the daily problems it causes linger on into the Twenty-First Century. There seems no way to kill the Color Temperature beast and the problems and misconceptions it causes. At the very least, as a sort of warning we should adopt the convention of using an image of a button hook in place of K for all color temperatures. (Buggy whips and button hooks were the two universal symbols of obsolescence in the previous century.) Thus we would write 4500? instead of 4500K, being more expressive of the actual situation.
Privacy Policy and E-Mail
Go to C F Systems Home Page