I have to start this article with a disclaimer: I'm going to show you some photographs of mine that should never have seen the light of day, let alone be posted on my website. Because we're talking about exposure - bad exposure - the best way to do it is to show you a few of mine. No laughing, OK?
A histogram is a chart that shows the amount of a specific tone that exists in the scene. I'll assume we're talking photoshop histograms, but most of the cameras I've seen present the information pretty much the same way. If you look at the histogram above, you can see that there is a predominance of tones that are a bit darker than middle grey, along with a lot of bright/white tones. There are a few spots in the image that are pure black and a very small number of spots that are pure white. In photoshop, there are sliders at the bottom of the histogram that allow you to adjust (i.e. "stretch") the tonal range of the image by dragging them around. Once you learn to read a histogram, you'll be able to very (very!) quickly assess an image's overall light range - and that's what this tutorial is about.
When you're doing digital photography, there are 3 tonal spaces you need to worry about:
Your camera is able to record a range of tones that it will represent as dark to light (there's color, also, but let's imagine we're just talking black and white) Usually this is a subset of what actually exists in the world. We can't really tell because our visual system is heavily post-processed by our brains, which create the illusion for us of a virtually infinite continuous tonal range from extremely dark to extremely bright. Our cameras record a slice of that and we call that slice "exposure." A photo that is too heavily oriented toward the dark tones we call "underexposed" and one that is too heavily oriented toward the bright ones is "overexposed" (or "burned out"). The reason this puzzles many photographers is because, if you look straight at a reflection on a piece of chrome on a sunny day, your brain adjusts the tonal range of the scene so it looks OK. Your camera can't do that, so you have to decide where you want it to record its limited slice of tones.
Look again at the histogram above. It offers a full tonal range from deep blacks (just a few!) to perfect whites (a smidgeon!) and your brain is going to look at the photo with those tones and adjust what it tells you you're seeing so that it looks like a very full, rich image with lots of subtle greys and highlights.
(Mrs T., Image producing the histogram above)
Now, that's not really a black-to-white tonal range, because you're looking at it on a computer screen. I have no idea what your screen has done with the image. But I carefully placed the tonal range of the image to be someplace where it's likely to fool your brain into telling you that you're seeing a nice full-range black and white image.
That brings us to the next point! I can control the histogram/tonal range of the image at capture time (what the rest of this article is about) and I can control it at edit time, in photoshop, but what about when you see it? Basically, all I can do is try to make sure my source image is as well-positioned in the tonal range as I can make it, and hope. If your display is too bright, the fine details in the white cloth may be blown out. If your display is too dark, the subtle shadows by her knees may turn completely black. But I've gotten the image as far as I could, with what I have available. If I were a complete control freak, I would only produce output on a carefully calibrated printer (where I could match the tonal ranges exactly) so that I wouldn't have to worry about whether your monitor is brighter or darker than mine.
A brief word about contrast, and then we're done with the theory section of this article. When you take an image that fits within a tonal range, and you 'stretch' the tonal range, you've just increased the image's contrast. "Contrast" is the separation (or distance) between tones in the image. When you're editing an image and you enhance the contrast, what you're doing is pulling the whites up and pushing the blacks down and widening the image's tonal range. This is important because images that are underexposed often look flat - lacking in contrast - and that's because contrast and tonal range are just two different words we use for how our brain tricks us into interpreting an image. We'll take a look at some underexposed images in a few minutes and you'll see what I mean.
In the meantime just remember that the tonal ranges between the 3 elements of a digital photo (capture, edit, output) should pretty much be as close to the same as possible, or you're altering the way the image will look. It's perfectly OK to alter how an image will look, but do it deliberately not because of some accident in your exposure or post-work!
I frequent a website that has a "critique" forum, and a recent thread caught my eye. A photographer had posted a series of images for critique that were, frankly, not very good. But nobody seemed to be able to come up with an effective and simple way to explain it without the photographer's getting defensive. To me, the images looked underexposed and, therefore lacking contrast. So I double-checked by loading the images into photoshop, checking their histograms, and then I posted a brief explanation of why the pictures looked flat and dark, concluding "this is an unhappy histogram" and including a screenshot of the graph in question.
(Click to enlarge)
The image above is an unchanged (except for size) shot from a series I did last year. On the right hand side is the histogram from the same image. This is a pretty good example of an "unhappy histogram"! Why?
First off, I know a few things about the image. Namely, that the piece of cheesecloth she's wrapped with is white and the background she is lying on is black. So the histogram shows clearly that my photo is underexposed, because the bright points on the histogram do not even appear until about 1/3 of the way across the chart. What the histogram is saying is that there are no white areas in the scene at all. I could fix this in photoshop but the problem remains that this is a sub-optimal capture. By not getting the exposure right, I've wasted almost 1/3 of the tonal range that the camera is capable of recording. If I "stretch" the image using levels and curves in photoshop I run the risk that the image will start to look "blocky" in the areas where I do the most "stretching" - which, in this case, would be the bright spots on the model's backside right in the center of the photo. I'm probably better off not spending any more time on this image and investing my effort in getting the shot right, next time.
(Egg white - click to enlarge)
Here's another histogram, but it's actually not an unhappy histogram. The photo is of a white egg on a white matboard. The overall tonal range of the image is white, white, white, with a small blue-grey shadow cast by the egg. If you look on the histogram you will see that most of the tones cluster where they should, at the extreme "white" end of the chart, but they are not going "off the edge" of the chart. The photo is too white but it's not actually blown out except for in a very small area at the upper right of the egg. This is basically a "rotten photo" but not an incorrect exposure. Because all the tones are right on the ragged edge of being blown out - to the point where my monitor can't render them very well - it would make sense if I worked on a version that was slightly less bright, so that I had more tonal range to work with. While the image above is certainly bright, it's not very contrasty! And that means I'm throwing away a lot of my camera's capability to record detail.
(A photo of my S3's LCD showing its own histogram)
I apologize for the poor quality of the photo above but I had to use a cheap point-and-shoot to photograph the LCD on my DSLR under less-than-optimal lighting conditions. It's a miracle that it turned out as well as it did, frankly. This is the master histogram display from my S3 for the same image as we were looking at, above. I found this exercise to be a real learning experience, since the histogram the S3 gives doesn't look a whole lot like the histogram I got back from photoshop. Which do I trust more? Since photoshop is my final imaging "device" it's got to take priority. But if I tried to use the S3's histogram as the sole means of determining exposure I would get different results than I'd get with a light-meter. More importantly, you can barely see the egg at all in the LCD. Since the LCD is back-lit the whole thing just looked like a big white blob with a tiny grey blob. I know some photographers who use the LCD on their cameras to adjust their exposure on the fly - it may be that they are making things look better on the LCD but are either over- or under-exposing the final image.
What's interesting is that this image was metered with a studio flash-meter. Was the meter "off"? No, I don't think so. But it gave me a resulting image that was probably a bit too close to the ragged edge of contrast for my tastes. That would be exactly what I would want if I were shooting this scene using film but since I'm using digital capture and post-working in photoshop, I'm less concerned about shooting a perfect negative for film than I am getting a capture that I can manipulate to sit exactly where I want it in my tonal range. Let's talk about tonal ranges, shall we?
Ansel Adams popularized the "Zone System" - a notation for thinking about exposure by treating it as a series of gradations on a tonal range. The Zone System was groundbreaking for its time because it let photographers talk about how they wanted to expose an image and what they expected it to look like when it was printed. Thus, a photographer could say "I want this area of the shadows to come out with just a bit of detail but be nearly black, and this highlight here should be the brightest spot in the photo." I won't bore you by translating that into zone-ese, but you can see that we could hold a Zone System style discussion around a histogram. The place where there's complete black? That's all the way on the left. The place where shadows just begin to become visible? Just a bit to the right of that, etc. There's no need to break the histogram up into arbitrary "Zones" - luminance is a continuum and it always was in spite of what Adams said.
In Zone System land, a photographer would shoot a negative and plan how the negative's contrast range would fall within the contrast range that could be recorded on the paper. Digital photographers have exactly the same problem, really. Our cameras can capture a certain range, and our monitors and printers can render a different range, and we need to make everything fit within the appropriate boundaries. The histogram is a really important tool for getting that right.
(Shooting fish in a barrel - click to enlarge)
If the earlier histograms we were looking at were "unhappy histograms" the one above is about as happy as it can get. There are a couple of things to notice. Most important, to me, is that the entire tonal range of the scene falls neatly within the camera's recording range. That means that the brightest white and deepest black in the scene all "fit" into the camera, so there is no loss of detail at either end of the histogram. Look at the histogram in the image above and you'll see 3 blobs - the black sheet of paper, the white sheet of paper, and the larger blob in the middle that represents my backdrop sweep.
Of course for the actual print (or display) of this image, I want the whites to be white and the blacks to be black. So I can pull the image up in "Levels" and make it so with a couple of mouse-clicks:
(Black and White in Color)
By pulling the histogram in I've chopped out some of the highlights and made them pure white, and the shadows and made them pure black. That was an aesthetic decision. Another photographer might put the black/white points a little farther out, so there was still a bit of detail in the shadows or paper texture in the highlights. The neat part is that I get to decide exactly where I want the luminance values to fall in my scene, because when I did my initial capture I didn't lose any information and it's all there for me to play with. The Zone System photographers, working with film and chemistry, did exactly the same thing: by carefully controlling the exposure and the film's development to capture the full contrast range of the scene, they got a negative that allowed them to print the resulting image exactly where they wanted it to fall within the paper's range between black and white. What I did above with "Levels" would be the equivalent of a Zone System photographer shooting a negative that had a slightly "compressed" tonal range and then printing it on slightly contrasty paper to pull the range back out, without losing any detail from the negative in the process. For "negative detail" you can substitute "information" and for "tonal range" you can substitute "histogram" and you and Mr Adams are talking the same language.
Let's take a look at what happens when your image is outside of the range that the camera can record.
(F-11 Blown out!)
In this shot, you can see that the whites just went completely off the top of the histogram. That means that some amount of information - we can never know what, exactly - was not recorded by the camera. That's not just "white" that's "overexposed." Basically, this photo is beyond repair. But we can still learn a few things from it: look at the black blob on the histogram where the flashmeter registers. It's pretty black! Our flashmeter may not be a perfect stand-in for a black card, but it's not awful, either. We'll come back to why that's useful later.
(F-8 Way blown out)
In the interest of science, I opened my camera's aperture up one more stop (from F-11 to F-8) so it would move the image's recorded range another "Zone" toward white. Now, you can see in the histogram that the scene is blown out; the tones are completely at 100% white except for the flash-meter, which has moved up closer to middle grey.
What have we learned here? You can figure out where your white-point is by taking a couple of shots of a white object and bracketing around using the histogram. In the example above, I could look at the way the grey blob from the light-meter was moving toward brighter and brighter and conclude that what I needed was less light, not more.
(F-16, bang on)
That is exactly what I did to get the exposure above. You can see that the black light-meter has moved more into the darker part of the histogram (where it belongs; note it's not really black it's just a dark grey!). Meanwhile, the background has become visibly grey and the only spots of pure white in the scene are the highlight on the egg and the light-meter.
So by looking at the histogram I can tell that my image is falling right in the tonal range that my camera can capture, and I'm not losing information at the dark or the light end. The lesson I learned from this whole experiment is that my light-meter reads about 1/2 stop off where I want to place my digital captures. Now that I know that, I will save myself a great deal of trouble.
When an old-school film photographer was presented with a scene that was too contrasty, they used a trick that allowed them to produce a full tonal range negative in spite of the fact that the scene was too bright to be completely recorded on the film. Basically, what we'd do is underexpose the film slightly, which meant that the highlights would not "burn out" in the negative. This resulted in a negative that would be slightly "flat" if normally processed. But then we'd take it into the darkroom and over-develop the negative in its chemistry to "stretch" the tonal range until it was exactly where we wanted it to be.
You can do exactly the same thing with a digital camera. There are two ways. One is to use a camera that supports a "wide tonal range" option, like my Fuji S3. It's not an option I use very often but it's nice to know that it's there, just in case. What the "wide tonal range" option does is tells the camera to sample tones above and below the ones it normally would and to use them as the black/white points. In other word, it "compresses" the image's tonal range (making it less contrasty) when it captures it. Then you can go into photoshop and "stretch" the tonal range back out to be exactly where you want it. That's just like what the old zone system guys used to do, only it's all done with nice clean digital methods and no smelly chemicals. The other way of getting the tonal range exactly where you want it is to shoot "RAW" images. When you're shooting camera RAW files, you're getting a (mostly) unfiltered version of whatever your camera's imaging chip recorded when you mashed the button. When you're not shooting RAW, your camera takes the data from the chip and then pulls out the slice of the luminance range that matches the exposure settings and ISO speed you told it you wanted. Once you have the RAW image in photoshop, you can decide exactly where you want the luminance range to be, etc, etc. I think a lot of photographers that swear about how great RAW is just don't understand exposure - their light-meter is photoshop - which is just fine, if it works for them. Since I spent years thinking about exposure during my "zone system period" I find it's very easy to compute exposure and visualize my image in terms of the three tonal spaces that matter:
If you think about your exposure that way, it becomes easy to decide where you're going to worry about getting it right, and where you're going to let things slide.
(F-16, You can't write a tutorial without breaking some eggs)
If you ever want a fun lighting exercise, try photographing an egg on a white background! The potential for contrast problems (too much, or not enough!) is really high because you're dealing with white on white and everything is reflective. Of course, it's quite simple if you're willing to spend a little time calibrating your exposure against some known objects and making sure that everything falls into the right place in your tonal range.
Bellwether Farm , Morrisdale, Pennsylvania Nov 29, 2006