2:05 amOff
Evaluating exposure via the Histogram
As you know, the image displayed on the LCD on the back of your camera is a poor tool to judge appropriate exposure—it cannot replace looking at your images on a proper display in a light-controlled environment. The trick, it is said, is to judge the exposure using the histogram—ideally applying what is known as the "Expose to the right" technique.
Moreover, as you also know, the image displayed on the LCD is not the raw file, it is a JPEG preview generated using the various parameters configured in your camera (white balance, saturation, contrast, etc.) Therefore, if your settings are such that the white balance is completely off the mark, for example, the image displayed will have a huge color cast. It's okay, since you know you'll easily be able to correct it later, because you're shooting raw.
What is less understood, on the other hand, is that the histogram itself is also not based on the raw data, but rather on that same JPEG preview! If you want to be able to judge the exposure using the histogram, how can you do it if the histogram is not based on the raw file? You can't!
...Well, let me re-phrase that: you can, but only up to a certain point.
What you should hope to achieve is a properly "exposed to the right" image—that is to say, an image that is exposed as far to the right as possible without clipping the highlights. That's a good thing, as we know that highlights hold more information than shadows, and what is nice about raw files is that they hold more highlights than what is possible when shooting JPEG. But how are we to know how far we can go if we base our judgement on a histogram that was built around a JPEG?
The idea is to use a set of parameters in the camera that will render the flattest JPEG preview, so as to generate the most generous histogram possible—regardless of the fact that it won't really matter in the end, as you are shooting raw and these settings are irrelevant to the final development of the image. If we were, for example, to use highly saturated colors and contrast to generate the JPEG preview (and, as a corollary, the histogram), we would be fooled into thinking that certain channels would clip before they actually do. Here is what you can do to render the most generous histogram possible:
- Use a bigger color space. If the camera allows you to use Adobe RGB (or even bigger) instead of sRGB, this will considerably extend the range of the histogram.
- Use minimal saturation and contrast. With Canon cameras (there are similar parameters in other brands as well, I am just not familiar with them), use a Faithful or Neutral "Picture Style", which will turn all the switches down.
- Use a white balance that is "close enough" to the actual lighting conditions—even if you intend to use something very different later, for effect. If you pick a Daylight white balance and shoot under tungsten light, for example, you might be fooled into thinking that the red/green channels are clipping before they actually do, as the JPEG preview will have a huge yellow color cast.
Needless to say, this will render a fugly image on the LCD, that will look completely washed out. Don't worry, you shouldn't be judging color/tint/saturation on the LCD anyway—do it for the greater good of a more reliable histogram.
Remember that the raw file still manages to contain more information than even this better histogram can display (because you will process the file using a bigger color space still, ProPhoto RGB), but at least you'll be much closer to the real thing.