Reading and understanding the histogram may seem difficult at first. But just remember that it's only a graph. The left end of the scale (see below) represents black. This is the point where shadows lose all detail. RGB values are 0 (on a 256 shade scale where 0 is pure black and 255 is pure white); nothing but black is being recorded. Brightness increases toward the right until it reaches pure white at the far right edge. At this point highlights record as RGB values 255 - pure white - and there is no detail.
The brightness value of every pixel is distributed across the histogram's horizontal axis, from black (left) to white (right). But this alone doesn't tell us enough. The photographer needs to understand how much of the image falls into a particular brightness range. A snow scene, for example, will be largely white, with only a small percentage of the image containing darker values. To communicate the amount of image data in a particular range of brightness, the vertical axis is used.
This graphic above correlates to a simple bar graph where the horizontal axis could represent years, and the vertical could represent sales over those years. The histogram should be read the same way. The histogram displays two aspects of an image - the brightness, and the amount of the frame that contains that brightness. The hard part is interpreting the information in the histogram. Sometimes it can be misleading, if you don't understand it correctly.
For example, a very dark image will show most of the data in the dark (left) areas. And a very bright image, like the snow scene, will have a graph that is skewed to the right. It would be easy to think of the exposures represented by these histograms as wrong. One appears to be underexposed and the other overexposed. This is deceptive, because the histogram's vertical axis is relative. A large amount of dark pixels will shrink the relative height of the bars representing the highlights, making it look like there is little data in the bright areas of the image. The same is true for a very bright image - the large amount of bright pixels makes it look like there is little data in the shadows. This is compounded by the small size and low resolution of most digital camera LCDs.
3. Histogram Examples
Let's look at a couple of photos and their histograms to better understand:
The Dark Photo
The image below contains lots of data in the very dark to black range. This histogram (right) might be misunderstood to indicate there's no detail in the highlights. But there's actually plenty of information. |
 |
|
| |
And the opposite:
The Bright Photo
The histogram for this brightly lit scene makes it appear as if there is no data in the shadows, even though there's actually plenty of shadow detail. |
 |
|
| |
The two samples above represent the extreme ends of the histogram spectrum. A typical photo's histogram will fall somewhere between those above. Some will have bumps in the middle, some will have valleys, and some will be pretty flat. It depends on the tonality of the image. Every histogram is as unique as the image it represents.
That said, there is one constant. There is always some data loss (clipping) in the shadows and highlights. These appear as spikes on either end and are clearly visible in the two extreme samples above. In both situations, the data loss was acceptable, even expected, due to the image content. When the histogram spikes like this the photographer has to decide if the data loss is acceptable or if an adjustment should be made to retain detail. Once again - clipping in the histogram isn't necessarily bad. It only informs and allows the photographer to make an educated decision about how and whether to change their exposure.
4. Variations on the theme
The histogram above shows the combined, average brightness of all three color channels (RGB) of an image. Some cameras, and most editing software, allow the photographer to view a histogram for each individual color channel as well as the standard luminosity graph. This RGB histogram is interpreted in the same exact manner as the single histogram above, but it gives the photographer even more information to work with. With an RGB histogram, you might see that the red channel is getting clipped before the green and blue channels, and adjust your exposure accordingly. Since the color data is averaged on a luminosity histogram, it might appear that no data is being lost when one color is actually being clipped. Highly saturated colors, for example, may clip one channel way before two others.
Finally, most cameras display the histogram with an overlaid grid that represents the five-stop exposure range of the digital sensor. This provides a quick visual guide to help the photographer decide how much exposure adjustment is required to place the tones where desired.
5. No easy answers
A histogram, though easy enough to understand, is open to interpretation to such an extent that some practice is required to truly grasp it. A photographer that wants to be able to take full advantage of it needs to study their images and the associated histograms. With time, interpreting a histogram will become intuitive and instantaneous.
- end - | 
del.icio.us
