Setting the Foundation for Good Color
A color image usually needs a visual base that establishes the accuracy of the color. That is normally a neutral color balance against which the other colors in the image are compared and recognized as being correct. If the colors in the image that would be perceived as neutral, especially blacks and whites, are noticeably contaminated by a color cast the entire image will appear to be incorrect. The natural adaptation of the eye to varying light uses neutrals as the foundation. Therefore, it is best to establish the contrast and color balance of an image by first correcting the black and white point neutrals. This is easiest to do at the start, usually within the initial stages of raw conversion, and makes other needed color correction within a file easier to recognize and accomplish.
Having said all of the above, there are some images which will not have a black point or a white point, or neutral values at all. Some floral images and abstracts lack such values. This floral image has no values below 70 and the highest value is about 240. While some areas on the right pass through neutral no specific neutral value is a good target area. The bright spot on the left is not neutral but appears to be in the context of the image. Making that point precisely neutral proved to be a bad idea in that it removed a desirable color cast that contributes to the mood of the image. A desirable color cast is one that might be seen in something like a sunset image where the colors are not accurate, but the effect is more important to the photo than the accuracy of the color.
In images without an obvious or natural neutral such as the floral image, a sunset, or other photos of nature the best solution is to choose the light balance preset option that most closely matches how the image was made. Even that can be difficult, especially in a sunset. It is often simply best to start with something like "daylight" and then modify the color temperature and tint sliders to achieve the desired appearance. In many images I run the presets to eliminate odd tint shifts and then reintroduce any desired color. In many images simply using the white balance eyedropper on a possible neutral point can eliminate color problems that are difficult to analyze, even if you then modify the color temperature for a more desirable rendering.
So, the first image I show you would appear to destroy my precept. However, most images will have a black point (the deepest shadow in the image) or they will appear to be lacking contrast. Because of the way the human eye perceives color, areas of very dark tone often seen as neutral even when they are not. Cameras do not have this perceptive ability and simply capture what is there, or add their own biases. The resulting reproduction does not match what the human eye would naturally see so the image appears incorrect. Even a profiled camera and a custom white balance set on a raw file can have a color shift in the blacks, and commonly this shift is to the red, especially in older cameras. While profiling should minimize this, setting the black point in Photoshop or Elements can go a long way to improving the color of most images. It is important to realize that the reason we need to do this is that images do not exist outside of an environment. We already have a frame of reference for the brain to work within, so the print has to exist within that framework. It must therefore be correct relative to the reality we see it in.
Setting a black point, however, does not mean making the lowest tone in the image zero in all three channels. A neutral dark gray is sufficient to establish a color reference. Any point already at zero in any channel is not a good target. In a gray scale of tonal values anything below 12-15 (or higher with certain papers) should be considered black absent a specific consideration for an output device. Numbers can and will go lower than that, but it is not necessary to artificially drive them down to lower values as that will simply drag other dark values down, possibly minimizing detail. Knowing how the image will be used, or reproduced will influence how the values need to be addressed. In most cases the three channels can be simply equalized at the value of the lowest channel, and adjusting the composite to meet the demands of the output device. This is because the reproduction of the image on a monitor, or printed by any particular printer on any particular paper will have a different black point and transition to dark grays that needs to be considered.
This landscape suffered from aerial haze, shadows overly influenced by a blue color cast and an overall lack of contrast. Roll your mouse over the image to see the original capture. The finished image benefits greatly from control of neutrals putting depth and contrast back into the image. Very little correction was applied to the file beyond the neutral corrections, and where applied had more to do with aerial haze than color. It is also interesting how much apparent sharpness appears in an image once the contrast is enhanced as our perception of sharpness relies partially on edge contrast. The image now looks considerably more like it did to the eye, and certainly is more compelling. Look also at the three dimensional effect of the corrected image compared to the original.
Aim Points and Sharpening
Most images have a high neutral value, but a pure white point requires that an image have a specular highlight. A specular highlight is a light source or reflection of a light source. This will be perceived to be the brightest point in the image as nothing could be brighter. Therefore, when printed, the specular would be represented by paper white. It is possible to find images where these concepts are intentionally distorted, but most realistic images have bright whites short of pure white that can be set as neutrals. A value of 245 is a reasonable target. Speculars can not be used as neutral targets as you cannot be sure at what point any channel has clipped compared to the others. The trick in processing a raw file is to target the brightest printable high values when dealing with the whites slider rather than to simply preventing clipping. If you are controlling an existing file in Photoshop or Elements the same concepts apply, but there you are raising certain channel values to match each other. Whether to use the central or highest of the RGB values when correcting for white balance is another judgment call based on how the decision affects other values in the image.
Black and white point corrections are most easily accomplished using Levels. The input endpoints immediately under the histogram are moved appropriately in each channel to achieve an aim point within the image. This distributes the correction in a decreasing percentage toward the mid-tones so that the modifications are gradual and even. Setting dark and light neutrals has the effect of cleaning up the color in an image. The key is to have aim points in the image that make sense, which can be set as probes in Photoshop or simply referenced as best as possible with the Info pallet in Elements. Be sure your Eyedropper is set to a higher value than the Point Sample default with a 5x5 average being a good target for most images. Use the Threshold adjustment layer to discover the darkest and lightest points in the image and help you decide where to make your decisions, and then discard the Threshold layer. Be prepared to choose an alternative point in the image should your initial decision result in a less than satisfactory correction. Some images with strong colors moving into dark areas can be problematic and knowing how your output device handles these tones will influence your decisions.
A critical point must be made here relative to image sharpening. Most sharpening techniques, particularly unsharp masking, expand the contrast of edge values within the image to create the illusion of sharpening. Saving a file as a jpg also can enhance the contrast of an image and push black and white values apart. It is common to notice white values clipping with sharpening and file conversion so this must be considered in the image making process. Black values will also change but are usually less obvious. One solution for Photoshop users is to attentuate the sharpening layer using the BlendIf sliders to protect the top and bottom values in the image. This is not complicated to do once you are used to the technique. This example shows values below 30 and above 232 being protected, mid-tone values from 60 to 200 being fully sharpened, and transition areas from 30 to 60 and 200 to 232. Sharpening is restricted to the mid-tones. The actual values for an image will vary based on the image content.
Sharpening with "Smart Sharpen" is a valuable technique as the filter incorporates the equivalent of BlendIf sliders in the dialog in an easily understood set of sliders. After setting the sharpening parameters you can restrict the effect to the midtones where they are most effective by attentuating the effect on the highlights and shadows independently and with excellent control options. This makes Smart Sharpen a favorite sharpening tool for many. Even using Smart Sharpen you should remember that painting the results into the image with masking is an alternative to be considered.
Midtone Neutrals and Color Shifts
Mid-tone neutrals are often harder to find a target for, and setting the white balance of the overall image in the raw converter will settle that issue well enough. In an already processed file if you know you have a neutral area or target, simply match the RGB numbers in that area regardless of value as anything with equal RGB numbers is neutral by default. The only decision is which channel to use as the set point. Usually the green is a good decision, and curves is a better tool to use as you can target values other than the mid-point for correction. In Photoshop you can set a probe on the area, and with curves you can modify the red and blue channel numbers to more closely match the green channel. This can be done in Normal mode if it does not adversely affect the brightness of the image, or in color mode if that is an issue. Values within 5 points of each other, usually with the green channel centered between the red and blue values, usually appear comfortably neutral. Strictly matching the numbers precisely is not usually necessary.
Some images look better when the neutral values are not strictly neutral. Often mid-tone values where a slight warming or cooling of the values imparts a desired color shift to the image are more appealing. Neutral black and white point values in those cases usually help make the shift in the mid-tones appear more intentional and accurate rather than looking like an error. Warmer skin tones are more appealing than tones that result from strict neutral interpretations of mid-tone values. It is the end points that set the stage for visual comparison. Even when this is done, departure from a real neutral is usually minimal.
A color temperature shift is where the RGB channels spread slightly with the red and blue channels moving equally away from the green channel. Where 140 in all three channels may be the accurate neutral color numbers, 145/140/135 might be a desirable, slightly warm neutral alternative, or even 150/140/130. How far to go is your decision. The neutral eyedropper in Levels can be adjusted to impart something other than neutral such as a slight warm or cool cast. That can be used to manage consistent color changes to a midtone neutral or match colors from one image to another. To avoid shifting the black and white aim points apply BlendIf to protect the ends or lock high and low values with points on the curves to minimize or eliminate the shift. Holding both the Ctrl [Cmd] and Shift keys down and clicking on selected high and low values will set lock points on the curves in each of the three channels.
A tint shift is where the green channel value is not centered or close to centered between the red and blue channels in a neutral value and will likely be seen as a mistake. Where a temperature shift (warm/cool) is common to achieve a desired appearance, a tint shift imparts a green or magenta cast to the mid-tone values rather than a warm or cool appearance. This is similar to the Tint slider you will see in the raw converter basic panel for color correction. It is common to see slight magenta (low green) casts in many images, often in skin tones. Intentional color casts may include such shifts, however, and a sunset with high magenta values will certainly be more acceptable than one where the color has been neutralized and therefore the visual impact diminished. That would be a case where the foundation black neutrals may be very important to prove that the mid tone tint shift is intentional and desired. If the magenta tint would bleed heavily into the deep blacks it could be seen as a processing error instead. Occasionally a color shift in the black and/or white points of an image can be seen as desireable depending on how "realistic" the intention of the image. This goes to what is known as "color grading" where the appearance of the image is intentionally shifted for effect.
Assuming you may not have set neutrals with the white balance tool in a raw conversion, the Levels and Curves dialogs include eyedroppers that will automatically set points for you. While this would seem to be the easiest way to handle the problem, especially with a jpg image, be aware that the default values for the eyedroppers are 0, 128 and 255, which are not necessarily the values you want in your files, especially at the ends. If you double click on the eyedroppers you can set new default values, like 12, 128 and 245, but they may not be desirable for every file. For those reasons I do not recommend the eyedroppers as the primary method of setting values as there may not be specific points that should match those values, especially in the highlights. You can try the eyedroppers to see if you get good results, but do not consider them infallible. If you need to adjust them for every file you may as well use other methods. The auto button in the Levels and Curves dialogs use the eyedropper settings for their decisions. The Enhance Per Channel option is best for that. The assumptions that are made here are that the image you are correcting contains appropriate values for those corrections. Low contrast images will suffer a rather horrible fate from use of auto correction. It is always best to target specific values in the image for manual correction.
While the black and white eyedroppers are usually best set to a neutral value, the eyedroppers can be set to a bias if desired. A warm white of 245/240/235 for example. This could be called a warm neutral as there is no tint shift and would correct a high value by setting it as close as possible to those numbers. The middle eyedropper is a bit different. It attempts to set the value you click on to a value with a similar difference in color channel relationships while trying to maintain a consistent visual appearance to the image. There is considerably less change in the visual appearance with the middle eyedropper, except for the color change. Photoshop works in the background in the L*a*b color space, not in RGB and it is therefore possible for the engine to shift the a and b color channels and have a lesser effect on the Lightness values in the image.
Therefore, once you have set black and white points, the mid tone eyedropper is as close to a white balance tool as you can get outside of the raw converter. You can even click in various spots on the image and see the change, choosing another spot if you do not like the results. Set a new default with a warm or cool shift and the eyedropper will use that shift as its balance. Set the defaults to 135,130,125 instead of 128,128,128 and your click will impart a warm neutral balance to the point where you click. This can be valuable in matching colors in multiple images as you can set values to effect whatever balance exists in one file and apply it to another file. The manual solution is to match RGB numbers in curves to points in each image.
Before DSLR still cameras the video industry had the ability to set the color balance of their capture by aiming the camera at a white piece of paper and setting the capture to be equal in all three channels, establishing a neutral balance under the lighting conditions, and giving us the term "white balance". Technically, the process is setting a neutral balance based on the fact that a neutral target is the same value in all three color channels (RGB). With film we chose to use daylight film, or tungsten balanced film and then added color correction filters to handle the rest. Without expensive color meters and a lot of filters a lot of photographers just lived with color that was close enough if shooting transparency materials, or relied on the labs to correct negative capture at the time of printing. Unfortunately, that meant that most photographers did not bother to learn much about color correction.
One of the first things we should do in processing a digital file is to establish a white balance, or neutralize the out of balance colors in the image. With a DSLR you can set a color balance to match the lighting source, such as daylight, tungsten, etc., or shoot a white target and designate it to be the custom white balance for the subjects that follow. In ACR or Lightroom you do this by making the target image the default and matching your corrections to the remaining files. In the absence of a target or an appropriate value within the scene you may find that cycling through the preset values; daylight, shade, tungsten, etc., will get you where you need to be, or close enough that a simple creative modification is all that is needed. While this works well enough for many images, photographers interested in accurate color balance will want to go further.
An accurate reference target will guarantee you the best possible color. I suggest the X-Rite Passport as a good solution to the issue as it will provide both good white balance targets and fully profile the camera for accurate color rendering under almost any light source. You can't beat a profiled camera for accurate color and in conjunction with an accurate white balance most of your color issues are effectively eliminated. A less expensive alternative is the Whi_Bal neutral reference card which is a good choice for a color reference but does not allow you to make a camera profile. I prefer reflective targets to the lens cap style or filter like methods.
Roll over the image to the right to see the original capture without the camera profile correction. The improvements in the blues in particular is very noticeable. The subtle changes in the various color chips in the target represent changing the captured color tones to the "accurate" color tones based on the target. This is not arbitrary, but rather based upon recognized industry standards. This is profiling the camera and will improve your overall color and diminish necessary local correction in post production.
Raw vs Jpg Original Files
An image has no white balance at the time of capture. If you shoot raw you import the image into the raw converter or Lightroom and see the image based on the way the camera was set when the image was made. This is simply taken from the exif information and/or an embedded jpg preview and provides you with a starting image for viewing purposes only. You will need to set the white balance using presets, or better yet by using the white balance eyedropper and choosing a neutral target area. The corrections possible in a raw file are amazing, but selecting the best point in the image to establish the white balance without a target can take a little getting used to. Fortunately, you simply click on various parts of the image until you see what you want. Since raw files contain more color information as the image gets lighter I usually recommend starting with a high value (180 - 220) rather than trying to set neutrals based on shadow detail or blacks, but sometimes it works fine.
If you shoot a Jpg image the color will be determined by the relationship between the color of light illuminating the subject and the color balance setting in the camera. The white balance settings in your camera will establish the color of the file and you are essentially locked into that decision. A custom white balance is a better choice here as the analysis is against a known neutral value and other colors will be recorded against that setting. Settings such as daylight, tungsten, fluorescent, or custom Kelvin settings for temperature make assumptions about the quality and color of the light source and process the images accordingly. Auto White Balance (AWB) is asking the camera to simply make a guess based on the image content and is the least reliable method. Even with consistent lighting the colors of the images will vary as the content of the images change, so every image will be different. While auto white balance may work on individual images a fair amount of time, careful analysis of those images will reveal inconsistent neutral values, and the results will not make the best possible image. Obviously, the closer to correct a Jpg image is coming out of the camera the easier it will be to resolve any color issues. Corrections are possible but severly limited compared to the options in a raw file.
The reason for owning a target of a known value is something that is also not understood well. White shirts and white paper may appear as white as a patch on the X-Rite Passport, but with one important difference - white comes in flavors. Shirts and paper are not a known source, but can have excessive amounts of ultra-violet light reflecting from them. They may also be less than accurate in color balance, yielding a different yellow/blue balance than a source designed specifically for this purpose. Therefore, the results from fabrics and papers may be less than satisfactory under many light sources depending on the spectral characteristics of the light.
In Lightroom or ACR the easiest part of the process is white balance. Using the White Balance Eyedropper Tool you click on a neutral target in the image and the software does the heavy lifting. This should be done after removing any clipping from the image. Then if a custom profile is available in the Calibration tab I use that, or for an image where the visual impression of the image is more important than an accurate representation, cycling through the profiles that came with the camera are an appropriate technique. Do not just pick the one that provides the most intense color, but try to decide what appearance you intend to make of the image and choose the profile that matches your concept. Profiles can exaggerate some colors over others and the right profile choice will help you get closer to your result without more work later in the process. After the profile is chosen, double check the white balance as it may have changed slightly. A white balance target value for me is generally a high value neutral well below clipping, perhaps something between 180 and 240.
With the white balance set we have eliminated tint shifts due to the lighting from the image. This means we have a technically accurate image, but not necessarily what we want to see. We can modify the color temperature to improve the skin tones in a portrait, or otherwise warm or cool the overall image to suit our intentions. I am guilty of favoring a slight global warmth in most of my images as I see warmer images as being more appealing. Further modifications may be made in to local areas using Photoshop, but the basic white balance in the raw image is an important first step.
If you are shooting Jpg images or dealing with an imported image from a scanner or other source, you need to deal with white balance in a different way. Once I have opened an image in Photoshop I essentially start the process over again, which is the process you will need to use with non-raw images. First I set my Black point. In raw images I keep my contrast a little low and my black point a little shy of where I want it to ultimately land. I mentioned that sharpening, output and other factors influence the needed black point so I leave myself room for adjustments. With the image open in Photoshop, the neutral base for my image is the starting point.
A Threshold adjustment layer is helpful in finding the darkest part of the image. That is where you will want to set your darkest value point and most often is also where you want to set the dark neutral value as well. Check the white point as well, and then delete the Threshold layer. Using the noted points in the image you use the Levels controls to modify the file and push the values the desired points. This should provide you with the basic image contrast and color balance needed to allow you to make creative decisions with the color issues primarily resolved.
Visual decision making assumes that you have a profiled monitor capable of showing you accurate color or you are asking for trouble later in the process. Prices of high quality LCD monitors are dropping all the time, but good monitors are still expensive. A quality monitor capable of being profiled is essential. A monitor that was designed for gaming or movie viewing will need to be modified for photography. Most monitors are cool rather than neutral, and often way too bright and contrasty for image editing. Once profiled and tamed they are easier to work with and also are less of a strain on the eyes over long periods. Accepting the color on your monitor as being accurate is not a good idea unless you have taken steps to be sure that it is.
It is an important part of learning to process digital images that you have good reference images to compare to your own. Learning what a good, color cast free, appropriately processed image looks like will allow you over time to more easily recognize problems on the screen. Initially, you can use target numbers to help you get an image reigned in, but the creative side of imaging requires going beyond the numbers to making informed visual decisions. Most photographers will just not go that far. Professional photographers certainly should and serious makers of fine art images will quickly realize the value in better processed images. The visual impact and refinement of vision that is apparent in a well processed image can be easily seen by the more sophisticated viewer. Understanding the role that a neutral color base plays in image making will help you get a better result, and get it more easily.