Where the Bits Are: Linear vs. Nonlinear Data

Chapter 2 touched briefly on the fact that your eye and your camera respond to light levels in very different ways. If you double the amount of light being shined at your camera, then your camera will record a 200 percent increase in illumination. Because a camera's response to light is directly proportional to the intensity of the light source, we say that it has a linear response to light.

Your eyes don't work this way. Your eyes are much more responsive to subtle variations in bright and dark tones than they are to changes in midtones. In other words, their response to light is not directly proportional to the intensity of the light source in question. Thus, we say that your eyes have a nonlinear response to light.

Like your eyes, film has a nonlinear response to light. As you saw in Chapter 2, when you shoot with your camera in JPEG mode, a gamma correction curve is applied to your image to make its luminance response more closely resemble the nonlinear perception of your eyes. Because JPEG's gamma correction yields a light response that is similar to the response of film, any exposure habits that you have from shooting film will most likely still apply.

When you shoot in raw mode, though, things work very differently. When you shoot in raw mode, your camera records a lot more data for brighter tones than it does for shadow tones.

Twice the data in half the stops

As you've probably already noticed, many of Camera Raw's controls seem to do the same thing. With the Exposure slider, you can brighten or darken an image, but you can also brighten an image with the Brightness control and darken it with the Shadows slider. Understanding how your digital camera captures different tones will make it easier to understand which control to use for a particular adjustment.

If your digital camera uses 12 bits of data per pixel, then it's capable of representing and processing 4,096 different levels of brightness. In a digital camera, half of those 4,096 levels go toward recording the brightest stop, half of the remaining levels go to recording the next-brightest stop, half of what's left from that go into the next-brightest stop, and so on for the remaining stops that your camera can capture. There is a straight, linear halving of data for each stop.

By the time your camera gets down to the darkest stop (usually the shadow areas of an image), it may have only 64 levels left that it can use to represent your darkest shadow details (Figure 6.9). A less pessimistic way of looking at things is to say that if you expose for the highlights, then you ensure that your camera encodes your image with the maximum number of brightness levels. (We'll discuss how to perform this type of exposure in Chapter 7.)

Figure 6.9. Most of the data your camera captures goes to recording the brightest half of the image. Half of the remaining data then goes to recording the next stop, and so on. This means that your camera records substantially more information for the bright areas of your image than for the dark areas.

As you've already seen, where there's less image data, there's a greater chance that posterization and tone breaks will occur when you make edits and adjustments. Because the camera captures so much data for the brighter stops, if you expose to capture as much information as possible from these areas, you'll have a tremendous amount of data to work with when you edit. This means that you'll be able to make large adjustments without fear of posterization.

Conversely, if you underexpose, then you'll be capturing more data in the midtones and shadow parts of the camera's rangeareas that aren't represented by a large number of levelsand so you will have less editing latitude (Figure 6.10). When you brighten this underexposure, you'll almost certainly reveal noise that's been hiding in the shadow parts of your image.

Figure 6.10. The image on the left was slightly underexposed. Though the shadows aren't clipping and it has an acceptable range of contrast, it has very little data in the highlight areas, the areas where your camera captures the largest number of tones. The image on the right has a lot of highlight information, meaning that it's a data-rich image that can withstand a lot of editing and adjustment.

To speak of this using terms that you learned in Chapter 3: You've seen that when you make a Levels or Curves adjustment or use the exposure sliders in Camera Raw, you are compressing or expanding different parts of the data in your image. Since your camera captures so little data in the shadow areas, expanding the shadows is a very bad idea, as it will almost always lead to posterization (since there's not much data there in the first place). Expanding the data-rich highlights into the shadow areas poses far less posterization risk, and because there's so little shadow data there to begin with, there's very little risk that you'll be compressing data that's already there.

You don't necessarily have to overexpose your images to perform this kind of capture. If your camera has a good light meter, you'll be shooting with exposures that yield very good data. However, your camera's metering system is probably optimized to yield exposures that are best for gamma-corrected JPEG images. Thus, many meters often err on the side of slight underexposure. While your meter will often be right, it's important to keep an eye on what it's doing. We'll discuss this subject more in Chapter 7.

If you're coming from a film background, where you're used to underexposing to protect your shadow detail, you'll have to retrain yourself. Underexposing when shooting raw is a bad idea. With digital raw photography, you want to expose to capture as much highlight detail as you can. You'll correct for your shadows later.

MEASURING IN STOPS

If you have even cursory photographic experience, you've probably heard the term f-stop. There are two mechanical mechanisms inside your camerathe shutter and the aperturethat control the amount of light that strikes the image sensor. When you change the shutter speed from 1/125 of a second to 1/60 of a second, you double the amount of light that strikes the focal plane, because the shutter is kept open for twice as long. Similarly, when you change the aperture from f/8 to f/4, you double the amount of light that strikes the focal plane because the size of the aperture at f/4 is twice as large as at f/8. (Obviously, moving the other direction, from 1/125 to 1/500 or from f/8 to f/16, results in a halving of the light.)

Each of these doublings (or halvings) of light is referred to as a single stop, and you'll often hear photographers using the word stop as a measure of light. A simple way to think about f-stops is to remember that a smaller aperture stops more light from hitting the focal plane, as does a faster shutter speed.

For example, say you're shooting in a somewhat dark situation and your camera's light meter recommends an exposure of f4 at 1/30 of a second. That shutter speed is a little slow to be shooting hand-heldthere's a good chance that your image will be soft or blurry at 1/30 of a second. Let's assume that flash is inappropriate in this situation. Let's also assume that you have a piece of white cardboard with you, which you use as a reflector to bounce some light onto your subject. Now when you check your meter, your camera recommends f/4 at 1/60 of a secondthat's a shutter speed that's twice as fast as what the camera was recommending before. With your reflector, you've just added an entire stop's worth of light to your scene and so now have an exposure more suited to shooting handheld.

Every doubling of light can also be referred to as one exposure value, or EV. Photographers often use EV to denote over- or underexposure without having to concern themselves with specific aperture and shutter speed values. So if two photos differ in exposure by 1 EV, then one of them was overexposed by 1 stop, through a change in either shutter speed or aperture (or possibly, by a change of half a stop in both parameters).

The human eye has a dynamic range of 30 f-stops of lightfrom fully adapted night vision under starlight to brightest daylightabout a factor of 1 billion to 1. The typical digital camera (or film, for that matter) has a range of 8 to 10 f-stops of light (or 8 to 10 EV). Some natural scenes have a range of 12 f-stops (or EV) between the brightest highlights and deepest shadows. The fact that the dynamic range of camera technologies is so much smaller than that of your eye is one of the reasons that photography can be tricky. Your camera often can't record the full dynamic range of the scene, or what your eyes can see, so you have to make choices about which 8 to 10 f-stops you want to capture.

Reading it in the histogram

From what you've already learned about reading a histogram, you know that when you darken an image, the bars in your histogram shift toward the left. This happens because, after your edit, there are fewer bright values and more midtone and dark values. Similarly, if you brighten an image, the bars in your histogram shift toward the right.

If you've exposed an image to try to capture as much data in the highlights as possible, your histogram will have a distribution that's heavy on the right, like the one shown in Figure 6.10.

Such an image may appear washed out, or too bright. However, because you have so much image data, you can easily darken some of those captured levels to produce an image that looks better. In the resulting image, the data will appear to have been pushed down into the shadows (Figure 6.11).

Figure 6.11. After the right image in Figure 6.10 has been adjusted, the histogram shows tones that have shifted to the left. We've darkened some of the bright tones to place them in the middle and shadow areas.

With all of this in mind, some of the seeming overlap in Camera Raw's controls should make more sense, as you'll see in the next section.