Histograms

A histogram is a bar chart of a frequency distribution. As used in photography, a histogram shows the distribution of the brightness levels or tonalities in a scene or image, from the darkest black to the brightest white that can be recorded on the sensor of a digital camera. Here is an example:

Across the bottom of the chart we have the “x axis” running from zero (no signal) at the left (utter blackness) to pure whiteness at the right, usually represented by the value 255. This is the maximum value that can be contained in each “8-bit” color channel of a JPG image, and corresponds to the binary value 1111 1111.

The height of each bar (actually just a thin line) in the histogram is proportional to the number of pixels having the tonality of the bar’s position along the x-axis, that is, the frequency of occurrence of that tonality in the image.

In the histogram shown above, there is no bar at the zero point at the left end of the x-axis, which tells us that all the pixels were at least a tiny bit brighter than totally black. It is usually desirable to have the entire histogram within the range of the x-axis. When the histogram is bunched up at the left or right end of chart, it strongly suggests that some areas of the image are actually off the end of the chart – they have been totally under- or over-exposed. The histogram is said to be “clipped”. This is somewhat the case at the right end of the example given. It is also possible, however, that the photographer chose to overexpose this rather small number of pixels – perhaps they correspond to bright white lights or specular reflections in the image. What you need to avoid is a serious clumping of large numbers of pixels at one or both ends, because this suggests that many tonalities were outside the recordable brightness range with the exposure given.

With a single exposure, of course, if the brightness range of your scene exceeds the dynamic range of your film or sensor, then you have to sacrifice one end of the tonal range or the other, or even both, to a lesser degree. In such cases (with an immobile subject), you could elect to make multiple captures at different exposures to cover the entire brightness range of the scene, and then combine the resulting images in post-processing. To a limited extent, you may be able to achieve a similar result from a single RAW capture using two or more different brightness settings for conversion, then combining these multiple images in post.

Histograms shown on digital cameras do actually represent the tonal distribution of the JPG image according to the current camera settings (contrast, color saturation etc). If you are shooting in RAW, then you will need to convert to your working format (16-bit TIFF, for example) using the exact same settings if you want to obtain the same histogram, but there is no reason why you should, and you are entirely free to use different settings for conversion if you wish, and you usually will. For the camera histogram to match the potential of a RAW image as well as possible, set the camera contrast at its minimum, and select the largest color space the camera supports. For DSLR cameras, this will usually be “Adobe RGB”.

You can think of the histogram display of your camera as a sophisticated light-meter, that gives you are spot reading for every pixel in your scene. If you can adjust your exposure so that the right end of the histogram as close as possible to the maximum brightness value at the right end of the x-axis (without clipping), you will capture the maximum possible amount of information for the best possible potential image quality. Of course, this exposure level may be too much for the subject rendition desired, so that you will then have to adjust in post-processing, but this approach will minimize the noise level in your image for whatever ISO setting you chose to use.

The bit-depth of RAW images, of course, is usually 12 bits or more, but these numbers are easily approximated in the 8-bit histogram by simply ignoring the least-significant bits. Any 12-bit value from 1111 1111 0000 to 1111 1111 1111 will be truncated to 1111 1111 0000, for example, and represented in 8 bits as 1111 1111. In the same way, 12- or 14-bit RAW values can be converted to the 16 bits per channel of your working format just by adding insignificant zeroes to the right. It is important to retain all the significant bits from your RAW files to maintain the best possible image quality during post-processing. That’s why we prefer to use 16-bit formats such as TIFF for our working files.

Some digital cameras can display histograms of each of the three color channels, rather than a single histogram of the overall brightness levels. This is extremely useful, as if clipping occurs in just one or two channels, the resulting highlight may display an unexpected colour cast in the image rather than the pure white suggested by a simple brightness chart.