Histograms and Expose to the Right
The histogram is your friend. This is your basic exposure guide with lots of detail. True, you have that little meter in the camera that indicates +1 or whatever but the histogram has far more and useful information. Most cameras will show you the histogram. You may have the option of a live-view histogram(that’s very good) or you may not get to see the histogram until you actually take a shot. (Still good but not as good.)
(Click on a picture to see it larger without cropping)
Histograms are simply a plot of the (relative) number of pixels with a certain luminance. The graphs are plotted so that no luminance – black in other words – is on the left while white is on the right. The famous “middle gray” would be in the center.
The histogram above is for a picture that would seem to be a pretty good exposure. It doesn’t crowd up against either the right or left edges of the histogram. In photographer speak, the highlights are not “blown out” and the shadows are not “blocked up.” Doing either of those things means you lose detail in the part that is against the edge. That may be ok depending upon what you want to do. Usually it’s best not to lose “important” details.
Here’s a histogram for an overexposed picture:This picture would be washed out with few, if any details. Notice the left half of the histogram is empty.
And here’s a histogram from an under exposed picture. This picture would be very dark.
And, here’s a picture that has limited contrast – that is, the histogram is fairly narrow – the real world tends to look like this much of the time. The scene looks flat when you view it on the screen.
Now, you may conclude from all this that you want to use an exposure that centers the histogram. But, that turns out not to be true. The mantra, if you will, is “expose to the right.” You want the histogram to be over to the right side without going over the edge. It’s a fine line and takes a little practice to get it correct. One thing is the histogram on the back of your camera may not be exactly the histogram you see in your image processing software. The reason is the camera is basing its response on a small jpeg of the image. If you are shooting raw, the histogram can be slightly different.
OK, but why expose to the right?
The relationship of f stops to light is such that the amount of light hitting the sensor changes by a factor of two for a change of one f stop. That is, from f/2.8 to f/4 is one stop. The amount of light hitting the sensor drops by 1/2. Or, from f/22 to f/8 – three stops – the light increases by a factor of eight. That’s an exponential relationship. 2 to a power.
Now sensors, on the other hand, are linear devices. That is, there’s no difference between adding 4 photons than 40 or 400 in output. It’s basically a one to one relationship.
What this means in words is that, for a given camera, fully one half of the levels that can be recorded by one pixel in the sensor will fit in one f stop on the right side of the histogram. For every f stop decrease, only one half of the remaining levels are available.
Hopefully the picture to the left will make this clearer.
This shows EV – a relative f stop, and the number of levels available for that f stop in a camera with a 14-bit A/D converter. A/D what? OK, for now just remember that that converter is capable of distinguishing 16,384 levels in a single pixel. But here’s a key point: fully one half of those levels are in the highest f stop. Then, as you can see, one half of the remaining levels are in the next stop and so on. Note the fact that one half of the possible levels or information are in the top f stop doesn’t depend on the number of bits in the A/D converter. It’s always going to be one half of the total.
In this case, I assumed the camera/lens combination was capable of a dynamic range – basically the distance in f stops between black and white – of six f stops. Your camera, if newer, will almost certainly do better than that. And, again, that doesn’t change anything for this discussion.
Here’s our histogram with the number of levels per f stop imposed – the one I said was “a good exposure.”
So,while the sensor could record 16,384 levels only 8,192 are being used by this exposure. The photographer “gave up” over 8000 levels. Notice the left side. There only 256 levels are available to record details in the shadows. Not good. Trying to recover the detail in the shadows only exaggerates noise in that area.
Here’s a picture Exposed To The Right.
Now the photographer is taking advantage of those levels unused previously. And, there should be more detail in the shadows should the photog want to develop them. Basically there’s more room to edit the picture. Keep in mind this applies to raw format. If shooting jpeg, you have much less headroom to edit the file and, therefore, want to make sure your exposure is more centered. Since jpeg conversion throws away information – it’s a “lossy algorithm” – some information is lost forever in the camera.
How about a ‘real world’ example of how this works.
Here’s the same scenes exposed to the right and underexposed (Yeah, OK, way underexposed to make my point more obvious.)
Now, it’s true you can adjust or “correct” the underexposure using Lightroom or some other image processing software. Check out the corrected version.
At this point you might be saying “All this to show me two pictures with no difference? ‘Sup with that?” Well, recall that we started out on this quest with the statement that we could do better, using noise as our figure of merit, if we were to expose to the right. So here’s a 100% crop of both the “correct” and “adjusted” images. On the left, is a 100% crop view of the corrected version while on the right – clever of me isn’t it? – is the exposed to the right version. Hopefully you see at a glance the problem The left version has a lot of noise that’s not there in the Exposed to the Right version. Why? Because the software is trying to represent the detail with only 256 levels of information. To bring up the exposure just exaggerates the noise that is there. Note that both shots were taken at the same ISO – 6400 so I’m not just showing you noise induced by the sensor and A/D converter.
So, bottom line is adjust your exposure such that the histogram is nudging but not hitting the right hand side. That way, you will maximize the headroom available for editing the picture. As I said, it takes practice. I would also mention that software programs like Lightroom, Photoshop and others can recover some amount of blown out highlights but you don’t want to depend on that too much.
You can also find a few folks on the web – but I think it’s a minority – that don’t agree with this idea. Here’s some references so you can start to decide for yourself. Have fun.
- Michael Reichmann, Maximizing S/N Ratio in Digital Photography
This is, so far as I know, where the name came from. The idea stems from Thomas Knoll, the person who first developed Photoshop
- Bruce Fraser, Raw Capture, Linear Gamma and Exposure.
- Martin Evening, The Adobe Photoshop Lightroom 3 Book, Adobe Press, page 275
- A/D Converters: See Discussion on Canon’s website