Color correction

The process for adjusting color to suit an aesthetic taste and the process for correcting faults in the color of an image are typically exactly the same.

For these examples we will need to look at specific software implementations, so as usual the interfaces and function availabaility will change between software packages.

Note: In photoshop in particular remember that use of any of the mentioned eyedropper tools will undo manual changes to the settings. So if you are going to use them then it's best to use them first and then do manual changes of the settings.

White balance

The subject of white balance can be difficult to explain because it involves trying to understand not how humans see the world but how *machines* see the world. It's none the less important to touch upon it since it is the cause of many color inaccuracies.

The concise explanation for white balance is that what the human eye perceieves as "white" light is actually made up of the entire spectrum and is not made up of different parts of the spectrum evenly. Some white light might have a bit more of the higher or lower wavelengths resulting is more reddish or bluish tones when seen by the narrower ranges used by cameras.

What hue the light shifts towards is dictated by the light source itself, not the surface it reflects off of, since the intensity of the light source is the primary influence on the dominant wavelength within the light.

9000 K

5000 K

3000 K

400 nm

700 nm

550 nm

In the above image you see a simple graph that relates light intensity and spectrum wavelength (in nannometers). You can see that lights on what we would consider the lower end of intensity (for house lights and such) "burn" at an intensity of 3000 K (Kelvin) which eminates primarily in the cooler end of the spectrum. Other lights around might be rated at temperatures around 9000 K which eminate light at wavelengths around 700 nm.

Just note that our use of the word "intensity" to describe the temperature of a light's color does not equate to the *brightness* of a light. They are different. It's perfectly possible to have a high-temperature, low brightness light.

There are actually a multitude of ways to alleviate these effects but these are probably the two quickest "quick" fixes one could apply for color correction...

Select a "Levels" adjustment from the "Layers > Adjustment Layers", and in the window that pops up...

  1. Select the icon of the white eyedropper.
  2. On your canvas (make sure you have the right layer selected if there are more than one) Left-Click an area that you know to be white within the real space (but that may not currently appear white in the image).
  3. Repeat with the black eyedropper, using it to click a black point.
  4. If more precision is needed, use the grey eyedropper tool to select a "neutral" tone. This means one that is not necessarily grey but one that has a medium brightness.

Usually you will see large changes with the use of the white dropper alone. Don't be afraid to use it alone if you are happy with the results.

Alternately, with the "Color balance" adjustment option...

  1. With your own eyes identify the hue your image has too much of (for example "my image is too blue").
  2. In the color balance dialogue adjust the slider of the hue that is too prominant within your image.
  3. Since you can only adjust one level at a time it's likely the first adjustment sent it too far into another range (correcting too much blue might leave too much red). So instead of reverting the change don't be afraid to adjust the other sliders as well until the image looks correct.

Truth be told most any of the adjustment types on this page could be used to fix the white balance of an image. What's best is to understand what each of them do so that you can apply the best adjustment to the image that is best suited to it whether it be for the sake of white balance adjustment or artistic expression.

As such let's look at them each more in-depth.

Curves & Histograms

A Histogram graph displays the prominance of a range of values within an entire image and / or selection.

There are several plug-ins, filters, effects, and adjustment functions within various paint programs that make use of what is known as a "Histoogram". The strict definition of a histogram is that it shows the distribution of data. In graphic design specifically they are shows the amount or prevelence of a particular property of an entire image.

Reading histograms

For instance we can use a histogram to show the prevelence of brightness in an image. You can just think of each small "bar" of the graph representing the number of pixels that the image contrains. There are two sets of images below, each has a source image, and the histogram produced by that image (within Adobe Photoshop).

Keep in mind that for these examples we'll be displaying *Brightness* within the histrograms.

Here we see a "black and white" image that is mostly grey tones. It has few black tones but a quite a few white tones. The histogram above reflects that. From left to right we see the distribution of dark to light tones. Because the majority of the graph rises near the center we can surmise most of the image will be made up of neutral grey values.
Now the histogram has changed to reflect the lowered brightness of the second image. Notice how the majority of the graph has shifted to the left side. This shows that while the level of absolute black pixels in the image are low, the very dark and near-black tones are more prevelent, and the pixels with lighter tones have decreased in number.

But brightness does not have to be the only property shown by the graph. It can show the amount of individual color channels, red, green, and blue, or it can show saturation as well. Theoretically any *single* property can be used. This just depends on your chosen program.

So how do we go about adjusting an image with this information?

Adjusting images with histograms

In the previous images you probably noticed two parts of the graph that weren't mentioned. The vertical greyscale bar and the diagonal line that cuts through each graph (the line is white on this particular graph).

At their core the horizontal and vertical bars (the ones that fade from black to white) represent the input and output values for any given value. The diagonal lines represents how YOU can modify the value between input and output. This is crucial to understanding the strength of curve adjustment.

  • The horizontal is the INPUT value.
  • The vertical is the OUTPOUT value.
  • The adjustable line represents re-assignment of the input value at the horizontal position to the output value at the vertical position.

This means a straight, diagonal line will always represent an un-altered image since the input value (along the bottom) will be the same as the output value (along the side).

So what happens when we are able to modify the line that represents input/output point?

Here is the original image we looked ato. The line has simply been grabbed (making a control point in the middle of the line) and pulled down. Most programs use this type of control-point and curved line system. Note how the histogram information remains the same as in the first image but the image itself is different because of the adjusted line.

Let's look at another example. In the next image we've adjusted the control line so that the contrast has been increased. Dark grey tones have been shifted to black and some of the light grey tones have been shifted to white.

Again we see the graph displaying the original information, but with the adjustment line modified to a steeper slope than the default diagonal. By bringing the black and white points closer together (making dark greys black and very light greys white) we increase the contrast greatly. In this case we did not click to add any new points we only moved the two that were at the ends of the line to begin with.

So is it possible to color correct with the histogram when these examples have shown mainly changes in brightness values? Yes, it will take a few more steps than a simple "levels" adjustment but can allow for more control. Mainly you will just make use of the color drop down. By default there should be an option that says either "Master" or "RGB" that you can click on and select one of the main color channels from. So if you wanted to lower the brightness of the blue channel it would simply be a matter of selecting "blue" from the channel drop down. At that point your adjustment of the median line would only affect the blue channel.

Finally, it should be noted that not every histogram will follow the format shown here. Photoshop for instance, by default now, might display the greyscale bars inverted so that the white tips are towards the lower left instead of the upper right. This does not change how the histogram works.

The following video relates to videogrophy, but the idea of the histogram can apply just as much to moving images as still images.

Brightness & Contrast

For the most part brightness and contrast adjustments within paint programs will be the most self-explanatory functions you can use. There are a few things to keep in mind about them though.

While most programs will use a built in curving algorithm (like that explored in the histogram adjustment above) some brightness and contrast sliders may apply in a linear fashion. But what does that mean?

Let's look at a random spread of numbers.

0, 1, 2, 3, 4, 5, 6, 7, 8, 9

Let's pretend these are brightness tones for individual pixels. It doesn't matter what the image is. Now let's apply a linear shift that increase the numbers by 2 (as if the pixels were being made brighter by 20%).

3, 4, 5, 6, 7, 8, 9, 9, 9, 9

Notice two things...
1. We no longer have any 0's or 1's.
2. We have 4 values of "9".

If these numbers represented pixels on an image this means we would no longer have any black pixels (only dark grey) and would have many more pure white pixels than we used to.

This is essentially another form of destructive editing. We've sacarificed black tones, "blown out" the higher values to white, and lost information in the image as a result.

Adobe Photoshop specifically will have the "use legacy" checkbox option. This is because the algorithm was changed to be adaptive to the image and avoid this exact problem. You'll want to leave it off for most operations.

Even if a program doesn't use linear shifting the threat of information loss is always there. This is just an inherent part of brightness/contrast adjustment. So always be aware of how much information you are sacrificing for your image. And if you simply want to change the brightness and / or contrast consider using the curves function above in place of a simple brightness/contrast adjustment as it leaves (somewhat) more information intact to revert changes at a later date.

Levels

In this context, the "levels" are similar to adjust the black and white points of an image. That means your primary method of control is adjusting when a one value (and everything below it) is considered black and when another value (and everything above it) is considered white. Those values are then applied to the image.

For the levels adjustment, if there is a "grey point" eyedropper available, then be aware it's not necessarily asking for the color grey as it it a neutral tone. Colors that are neighther too undersaturated or oversaturated, dim or bright.

The movement of the black and white points are fairly easy to understand, they involve shifting how dark or bright a tone should be before being converted to pure black or white, but the "input" and "output" fields deserve some explanation.

Essentially the "input" positions dictate where the "output" fields will begin. This is why you can put the white point of the input slider to the middle of the range and upon moving down the white point of the output you'll see parts that were white before turn grey. What was being sent in as the whitest part of the range is now grey. This particular example would be akin to dragging down the right-most handle of a histogram curve into the grey tones.

Via the adobe web resource:

"By default, the Output sliders are at level 0, where the pixels are black, and level 255, where the pixels are white. With the Output sliders in the default positions, moving the black input slider maps the pixel value to level 0 and moving the white point slider maps the pixel value to level 255. The remaining levels are redistributed between levels 0 and 255. This redistribution increases the tonal range of the image, in effect increasing the overall contrast of the image."

Exposure / Gamma

These settings are really only useful for 16 & 32 bit images. Exposure settings typically seek to emulate the f-stop of a camera and thereby change the amount of light in a scene after the picture has already been taken.

This option is best explained by the discussion of 32 bit images in the "Image file preparation" section.

The Channel Mixer

The channel mixer will be one of the first adjustment algorithms used when you just intend to color-correct an image. This is an adjustment type that allows you to fine tune the amount of one channel of color, red, green, or blue, present in the pre-existing colors of the image.

Note: the way the channel mixer works within Adobe Photoshop is nearly identical to the way it works in Adobe After Effects. If you can color correct still images, you can color correct video as well.

There is one singular concept to realize when using this tool. When selecting a color channel from the main drop down you ARE NOT selecting which colory you want to edit. You ARE actually selecting the amount of that color in the sliders below it.