Can you Evaluate Exposure Using the In-camera Histogram?

They say that "a histogram is a graphical representation of the pixels exposed in your image" or "when judging exposure, the primary areas of the histogram to be concerned with are the right and left edges", and illustrate it with something like fig.1 below:

Typical Histogram legend

Figure 1.

The problems with presentations like the one above is that they tell us nothing about exposure per se, they aren't in any familiar photographic terms, and that's not even mentioning that "very dark" or "very light" is about brightness, not about exposure.

In an attempt to put some numbers to those graphs, we took a series of shots, spanning over 10 1/3 stops, decreasing the exposure (increasing the shutter speed) by 1/3 of a stop for each consecutive shot. The exposure was set in such a way that the entire first shot "blinked" on the camera LCD, while the next shot, exposed 1/3 of a stop lower, did not.

The target was a simple halogen bulb in a reflector (you can use a LED panel, just make sure it doesn't flicker), diffused with 3 layers of heat-resistant white translucent fabric (the type used for softboxes, umbrellas, shooting tents...), placed approx. 2' in front of the light. The lens hood nearly touched the diffuser, the lens was focused on infinity.

Please keep in mind that the histograms your camera displays are from JPEGs, even when you are shooting RAW. All digital cameras always capture RAW data, and render a JPEG based on that data and camera settings. If you are shooting JPEG, RAW data will be discarded. If you are shooting RAW, JPEG will be embedded into the RAW file and used for previews and histograms.

Nikon Camera Histograms

Figure 2. From left to right: #2640, #2645, #2649, #2654, #2666.

Top row: Luma histograms of respective shots; bottom row: Luma+RGB histograms.

Please click to zoom in

As you can see on fig.2 above, the histogram field is divided into 4 areas, and we have five vertical lines: left wall, 3 dividers, and right wall. Traversing this figure left to right:

  • The histogram for the leftmost shot on the figure above, #2640, is very close to the right wall, and the entire image blinks on the LCD screen.
  • The histogram for shot #2645 is aligned with the first divider, thus the first area includes 1 2/3 stops.
  • The histogram for shot #2649 is aligned with the middle divider, and thus the second area includes 1 1/3 stops.
  • The histogram for shot #2654 is aligned with the last divider, and thus the third area includes 1 2/3 stops.
  • The above 3 areas taken together contain 4 2/3 stops.
  • The histogram for shot #2666 is very close to the left wall, and thus the last area includes 4 identifiable stops, and the last one of those is barely separated from the left wall.
Calibrated Camera Histogram

Figure 3. Nikon camera histogram calibrated in stops

A histogram allows one to differentiate slightly less than 9 stops (4 2/3 in the first three areas plus 4 in the last one, a rather natural choice for 8-bit JPEGs).

Let's see how statements like "when judging exposure, the primary areas of the histogram to be concerned with are the right and left edges" hold.

The last histogram area, the shadows, should contain everything but what the three other areas contain. Given that the dynamic range of the camera "in RAW" is 10+ stops, it's more than all other areas combined, at least 10 - 4 2/3 = 5 1/3 stops, represented in a very compressed form, making it complete guesswork to determine the clipping of the shadows, the presence of details in the shadows, shadow "pushability". It doesn't look like examining the shadows of the histogram tells a whole lot to a RAW shooter.

On the same note, a histogram is also not very useful for evaluating the highlights in RAW - while the in-camera histogram for #2640 is dangerously close to the right wall, indicating essentially no headroom (not to mention that the whole frame is one solid "flashing area", indicating overexposure), the RAW has a headroom of slightly more than 1/2 of a stop before highlight clipping; at 1/2 only very few pixels are clipped, see fig. 4.

FastRawViewer. Exposure correction 0.5 EV. OverExposure indication

Figure 4. The shot #2640 opened in FastRawViewer: without any adjustments (left), with +0.5 EV "exposure correction" applied (right)

Again, based on the camera's histogram, we can't say by how much we can increase the exposure for RAW, we can only guesstimate. And that's already for the simplest of cases, the case of a neutral subject. With a yellow daffodil, red rose, or blue sky it is much more complicated, please have a look at How to Trash a Good Shot in One Step...

So, those left and right edges of the histogram are not exactly suitable for the purposes of judging how well the RAW is exposed; that is, they don't tell us how much can be recovered from highlights and shadows, and what will be the penalty for such recovery.

The other issue that makes it incorrect to refer to a histogram as "a graphical representation of the pixels exposed in your image" is that a histogram changes significantly with changes in the camera settings such as contrast, picture style, brightness, white balance, etc. Here is why. Those camera settings change the in-camera JPEG rendering, that includes the JPEG embedded as a preview into the RAW file. The histogram is derived from that embedded preview, and as that preview changes, so does the histogram.

Even if you've carefully calibrated your in-camera histogram, changing one of the settings, or, worse, setting it to Auto, throws the entire calibration off. Auto dynamic range modes, auto contrast, D-Lightning, etc. affect the relation of the histogram to the exposure by quite unpredictable factors. Here are a few examples.

We picked a couple of RAW shots from the above series, and rendered them in the camera, changing one of the settings. The same RAW data, rendered differently because of the differences in camera settings, resulted in very different embedded JPEGs, and, consequently, in very different histograms:

2645 NL

Figure 5. Changing Picture Style from Standard (top row) to Neutral (bottom row),

the histograms moved to the left, yet the exposure, as well as the underlying RAW data, are the same

2649 with and without WB

Figure 6. Changing white balance, exposure remained the same, the underlying RAW data is the same,

yet the histogram is completely different

2645, Contrast 2

Figure 7. Changing Contrast setting in the camera,

histograms moved to the right, yet the exposure, as well as the underlying RAW data, are the same

In conclusion of the first part of this 3-part study, we can say that for the neutral subjects:

  • if you are shooting JPEGs, the histogram is useful, though it lacks resolution in the shadows, and of course it has little to do with photographic exposure;
  • if you are shooting RAW, it is worse - the histogram misleads when it comes to exposure of the highlights and shadows.

On a side note, the middle divider on fig.3 roughly corresponds to the brightness of the middle of Zone V.

Comments

From some tests, i wonder that camera histogram is a linear scale of JPG view with 256 on the right and128 at center (-1ev). The first squater reach 64 (-3ev) ...
With Raw digger we can see a Log scale as in the "zone system"  : the different scale create a compression of left part of histogram, as you say.
To see that, viez histograms of lin or log grey scales.

I try to understand : "Zone V" is 2,5ev below white (each zone correspond to 1/2 ev), it correspond to 18% grey scale ...
For a 128 grey, the histogram spike is at the middle of the lightroom histogram, and is only 1 ev below white !  (and Lightroom histogram seams to be equal to Nikon histogram)
So, on fig3, the -3ev (256/8) line would be very on the left of camera histogram .... ?
To confirm, we have to create a Log and a Lin grey scales, and see them with photoshop or lightroom ...
(excuse me for my badly english)
You can  anser me to gilles.toutou@gmail.com
 
 

Dear Sir:

"Zone V is 2,5ev" and "each zone correspond to 1/2 ev" - how so? Please use support@rawdigger.com for this discussion.

Add new comment

Simple HTML

  • Web page addresses and e-mail addresses turn into links automatically.
  • Allowed HTML tags: <a> <em> <strong> <cite> <blockquote> <code> <ul> <ol> <li> <dl> <dt> <dd> <img>
  • Lines and paragraphs break automatically.
Subscribe to this blog updates