Posted on October 17, 2017
One of the greatest benefits digital photography has over film is the ability it provides to check an image’s exposure at capture (when you can do something about it). But as photographers, we rely so much on our eyes that it’s sometimes difficult to accept that they’re not always right. We take a picture, look at it on the LCD, and decide whether or not it’s perfect without even considering that there might be a better way. But where exposure is concerned, there is indeed a better way.
That’s because every picture you click, whether raw or jpeg (but especially raw), contains more shadow and highlight information than the jpeg that appears on your camera’s LCD preview screen can reveal. Fortunately, our digital cameras give us a tool that tells us about that missing exposure info: the histogram.
A histogram is a graph of the tones in an image. While I imagine that any graph has the potential to evoke high school science trauma flashbacks, a histogram is really quite simple, simple enough to be read and interpreted in the blink of an eye. And not only is your histogram easy to read, it’s really the most reliable source of exposure feedback.
When an image is captured on a digital sensor, your camera’s “brain” samples each photosite (the individual pixels in the megapixel number used to measure sensor resolution), determining a brightness value that ranges from 0 (black) to 255 (white). Every brightness value from 1 to 254 is a shade of gray—the higher a photosite’s number, the brighter its tone.
Armed with the brightness values for each photosite in the image, the camera is ready to build the image’s histogram. The horizontal axis of the histogram has 256 discrete columns (0-255), one for each possible brightness value, with the 0/black column on the far left, and the 255/white column on the far right (they don’t display as individual columns because they’re crammed so close together).
Despite millions of photosites to sample, your camera builds a new histogram for each image instantly, quickly adding each photosite’s brightness value to its corresponding column on the histogram, like stacking poker chips—the more photosites of a particular brightness value, the higher its corresponding column will spike.
Reading a histogram
The version of a picture that displays on your camera’s LCD is great for checking the composition, but the range of tones you can see in your LCD preview image varies with many factors, such as the camera’s LCD brightness setting and the amount of ambient light striking the LCD. Most important, because there’s more information in captured than the LCD preview can show even in the best conditions, you’ll never know how much recoverable data exists in the extreme shadows and highlights by relying on the LCD preview.
It’s human nature to try to expose a scene so the camera’s LCD image looks good, but an extreme dynamic range image that looks good on the LCD will likely have unusable highlights or shadows. As counterintuitive as that feels, exposing an image enough to reveal detail in the darkest shadows brightens the entire scene (not just the shadows), likely pushing the image’s highlights to unrecoverable levels. And making an image dark enough on the LCD to salvage bright highlights darkens the entire scene, all but ensuring that the darkest shadows to be too black. In fact, a properly exposed extreme dynamic range scene (a scene with both bright highlights and dark shadows, such as a sunrise or sunset) will look awful on the LCD (dark shadows and bright highlights). The histogram provides the only reliable representation of the tones you captured (or, in your live-view LCD display or mirrorless electronic viewfinder, of the tones you’re about to capture).
There’s no such thing as a “perfect” histogram shape. Rather, the histogram’s shape is determined by the distribution of light in the scene, while the left/right distribution (whether the graph is skewed to the left or right) is a function of the amount of exposure you’ve chosen to give your image. The histogram graph’s height is irrelevant—information that appears cut off at the top of the histogram just means the graph isn’t tall enough to display all the photosites possessing that tone (or range of tones).
When checking an image’s histogram for exposure, your primary concern should be to ensure that the none of the tone data is cut off on the left (lost shadows) or right (lost highlights). If your histogram appears cut-off on the left side, shadow detail is so dark that it registers as black. Conversely, if your histogram appears cut off on the right side, highlight detail is so bright that it registers as white.
Managing a histogram
In a perfect world, when you see your histogram cut off on the left (everything cut off on the left is detail-less black), you simply increase the exposure until the histogram shifts right (brighter) far enough that no shadow data is cut off. And if you see your histogram is cut off on the right, you decrease the exposure until the histogram shifts left (darker) enough that no highlight detail is cut off. Problem solved.
But many scenes contain a broader range of light, from the darkest shadows to the brightest highlights, than the camera can handle. In these scenes you can blend multiple exposures that cover the entire range of tones, apply a graduated neutral density filter (to moderate the sky). When those options aren’t available or practical, I usually save the highlights and sacrifice the shadows.
While the general goal is to ensure that none of the tone data is cut off on the left or right side of the histogram, the exposure you choose for a scene is ultimately a creative choice that isn’t bound to the way the scene looks to your eye. Though I often expose my scenes to match the amount of light my eyes see, sometimes I decide to make the scene darker or brighter than what I see.
A picture is worth a thousand words
Trusting a histogram over a picture you can actually see requires a leap of faith. I can explain the concept until I’m blue in the face, but in my workshops the point doesn’t usually hit home without a demonstration. For example, the two Horseshoe Bend sunstar images below are from the same file—on the left is the way the picture looked when I captured it, along with its histogram; on the right is the same picture with just a few minutes of very basic processing in Lightroom and Photoshop (no plugins, blending, or any other elaborate processing).
If I’d have exposed this scene bright enough for the shadows to look good on my LCD (more like my eyes saw them), the highlights would have been hopelessly overexposed (white); if I’d have darkened my highlights enough to look good on my LCD, the shadows would have darkened to an unrecoverable black. I knew my best chance for capturing this high dynamic range scene with a single click was to ignore the LCD and trust the histogram.
Despite an image that didn’t look good at all on my LCD, the histogram on my Sony a7R II showed me that I’d captured virtually all of the scene’s shadows and most of its highlights. And because I captured this image in raw mode, I was confident that I had even more shadow and highlight information than my histogram indicated, a fact instantly confirmed with a rightward tug of Lightroom’s Shadows slider. With minimal processing effort, I was able to achieve final result you see here. (This is why landscape photographers are always begging for cameras with more dynamic range, and also why I use the Sony a7RII.)