Histograms and Using Your Head

If you saw a histogram like this one on the back of your camera what would you think? Mull it over while I yammer on about what a histogram is. A histogram is a graph representing the relative number of pixels of a particular tonal value and in the case above a color value as well. Great what the heck does that mean. For now we'll stick with tonal values for simplicity. Tonal value 0 would be all the way on the left and it is dark. Tonal value 255 would be all the way on the right and it is light. Tonal value 128 would be in the middle and is… well… medium. If the pixels in an image were all black there would be a line going strait up to the top of the histogram all the way on the left. If all the pixels were white that line would be at the right. If they were different values but close together there would be more of a mountain kind of shape in the histogram at what ever tonal value where they fell. The steeper the mountain the closer the pixels are in tone. The more gentle the slope the farther apart in tone the pixels are. Sometimes it is hard to envision how many pixels are actually representing what in a histogram. Especially a histogram like the one presented above. When looking at something like this it might be helpful to look at areas of the histogram like containers or drinking glasses with the containers all being the same height but having different widths. A very narrow glass doesn't take a whole lot to fill it up. Imagine a glass the width of a straw, a little bit of water fills it up to the top real quick. Now imagine a glass the width of your bathtub, it takes a whole lot of water relative to the straw to even be able to measure let alone reach the height of the little bit of water in the straw. So imagine that all of the pixels in your image are water in a can and the histogram above has two containers one is over towards the left and is is real skinny, the other is the rest of the histogram and is pretty darn wide. If start pouring your pixels into the skinny one on the left it is going to fill up really quick and most of your pixels are left in the can, when you dump the rest in to the container that is as wide as the rest of the histogram it's not going to be filled up very high relative to the skinny container on the left.

So back to what the histogram at the top of the article means and what would you do if you saw it on the back of your camera. Is it underexposed because there is a big mountain of pixels on the left and what appears to be a sprinkling of pixels over the rest of the histogram? Maybe it is. In this case it is not, it's perfect. Here is the image that is represented by the histogram above. This is where using your head comes in. Where is the mountain of dark pixels on the left coming from? Answer - from the black background, black hair, black wardrobe as well as the shadows on the skin, the sofa, etc. The reality is that most of the image has great tonal range, heck it uses up the rest of the histogram. You should see what happens when I turn the light with the red gel that is on the background off (more on that topic some other day).

The reason that I had to write this article is because I read so much crap regarding what a histogram “should” look like. Something along the lines of "It should be a mountain in the middle and slope off towards each end of the histogram". Bullshit! I dare you to expose this image so that that mountain on the left is in the middle, in fact if you move it up more than a bit I can guarantee that it will be crap, especially on digital. The exact opposite holds true if you are shooting on a white background or with a light source in the frame (on purpose) or with specular highlights (shiny stuff).

I have had more participants in my lighting workshops that were following crappy histogram advice, rules, whatever without using their head than I care to count. An example - a really nice guy and pretty good photographer that was using a high-end DSLR was getting consistently crappy images, I mean really crappy. When we took a look at what he was doing it was obvious. His process was something like this:

Set up white background and light it. Set up model and light him/her Meter light so that it is one stop over on the white background and right on for the model. Set camera to light meter reading, take test shot, and reduce camera exposure till mountain is near the middle of the histogram.

Let's take a look at what is going on here. He is taking that big spike on the right side that represents all of the things that are supposed to be pure white and moving them to the mid-tones and pushing all of the mid-tones and shadows (the important part) into murky digital sensor hell. But wait he shoots RAW so everything's fine, right? He brings them up in the RAW converter and pumps the exposure back up to where the background is white and viola. WRONG - yes it is an image but it is awful. Digital sensors are amazing at capturing shadow detail, so yes his mid-tones come back to being mid-tones but they have no contrast and tons of noise but he wasn't "blowing the highlights", you know the all important rule in digital photography, hogwash.

The bottom line is that you have to use your head. Get to know your tools, use the histogram it's one of the most awesome light meters out there but do not be a slave to it and stupid pet tricks that you read about how things should be. Some things are dark, somethings are white, some highlights have no detail and that is the way they should be. Backing off exposure until the specular highlights on a shiny piece of metal/latex/light source/white background/etc are not causing peaks on the right side of your histogram will generally give you noisy murky images even if you shoot RAW and crank it back up in post. Cranking up exposure so that things that are supposed to be dark or black come off of the left side of your histogram will most likely blow highlights that you do care about.

RB

blog comments powered by Disqus