A histogram is a chart that shows the distribution of pixels throughout the tonal range of an image. Its purpose is to help you identify irregularities. A tone curve or gamma curve is a tool that lets you make smooth adjustments to the tonal range of either the overall image or individual color channels.

Reading histograms

The histogram of a grayscale image contains 256 vertical lines (0 to 255), each representing a specific gray level. The heights are proportional to the number of pixels per gray level. For RGB images, a combined histogram indicates overall brightness. But you can also view each color channel separately.

An absence of pixels in a number of consecutive gray levels, as shown in the image, indicates tonal banding or posterization. This can happen if you manually set the contrast and highlight outside the range of the image. This is the original scan with no correction, as indicated in the tone curve graph by the straight diagonal line. Moving the curve to an extreme expands the shadows and highlights across a wider range causing posterization. Photos courtesy of Agfa

The distribution of pixels, especially at its extremities, provides a guide for tonal corrections. The top-left scanned image is low in contrast, having virtually no pixels at the black (0) and white (255) ends of the histogram. For now, ignore the little square graphs with the red lines. These are tone curves, which we will get back to later.

An absence of pixels in a number of consecutive gray levels, as shown in the image of the woman, indicates tonal banding or posterization. This can happen if you manually set the contrast and highlight outside the range of the image. Here is another reason I dislike manual settings — if you adjust the highlight and shadow settings of an image that has a wider tonal range than the settings you chose, clipping occurs. The histogram will show very high values at both ends. Shadow detail is forced or “clipped” to black, and highlights are “clipped” to white.

Scanners with automatic density control create internal histograms after a preview scan, from which they determine correct shadow and highlight settings. The final scan then captures the full tonal range without posterization or clipping.

An unevenly distributed histogram does not necessarily mean that the image is incorrect. Severe posterization, for instance, is sometimes used deliberately as a special effect. It will render the image to a flat black and white, like a silhouette. More importantly, there are images that, based on content, contain very few shadows and should not be tampered with. These are called high-key images. A low-key image has few highlights.

Using tone curves

One way of correcting histogram readings is to use tone curves. A single curve modifies overall brightness levels in color images. Or you can use separate tone curves to change RGB colors individually. The curve tool lets you make smooth changes across tonal ranges and is relatively easy to manipulate.

Normally, the horizontal axis shows the unchanged scanned image (input values) and the vertical axis shows the effect of the tone corrections (output values).

Let’s go back to the image of the woman. Assume this is the original scan with no correction. This is indicated in the tone curve graph by the straight diagonal line. Moving the curve to an extreme expands the shadows and highlights across a wider range causing posterization.

A more moderate adjustment (as shown in lower right image of the woman) — slightly lifting the three-quarter tones, and dropping the highlight quarter tones — improves the shadow detail without sacrificing highlight detail or producing disturbing posterization.

While there will always be times when you want the flexibility to manually adjust tone curves, my advice is to look for powerful scanning software that will let you make and store user-defined tone curves. Or better yet, image-processing software that will read a scan’s tonal values and make these adjustments for you automatically.

My thanks to Agfa for providing the images and diagrams.