"A "good" histogram doesn't have to stretch entirely from the left to right side to indicate proper overall exposure."
Histograms display a graphic representation of the exposure on a captured image. They provide highly accurate information, but their use is often misunderstood, or, worse yet, ignored.
How do they work?
Histograms are a graph showing brightness levels of pixels in a recorded image. The camera's processor locates each picture element (pixel) on a horizontal scale on the histogram, according to its relative brightness from 0 (pure black) to 255 (pure white).
Bright pixels are located toward the right of the graph; progressively darker pixels to the left side. Any pixels that are so bright that they'd reproduce as a pure white, or so dark that they'd simply show as a pure black with no detail, would be at the extreme right/left edges of a histogram.
Pixels of equal brightness are stacked vertically to create lines of varying heights. The result is a graph of very fine vertical lines that can appear as a smooth curve, a series of jagged lines, or a combination of both. The HEIGHT of the graph within a histogram says nothing about how bright or dark the scene is; it's strictly a reference to how many pixels in a scene were recorded at a given exposure level. Most users will interpret the important side-to-side dimension when evaluating histograms.
How to read a histogram: overall exposure brightness
Histograms accurately describe the brightness information from a specific image, but there are no "good" or "bad" histograms. Different subjects might generate completely different histograms, but both could be properly exposed.
For example, a white rose on a white background, if properly exposed, will generate a histogram pushed well to the right, because there are no shadows and very few, if any, mid-tones. In this example, virtually all the pixels in the image will show a fairly high brightness level in a properly exposed image.
Likewise, a black cup on a black background will display a histogram pushed to the far left, which normally would be a sign of underexposure, but in this case, in the actual scene, there are few mid-tones and no highlights. A properly exposed image of this type of scene would produce a histogram where the vast majority of pixels would be dark to near-black in brightness level. After all, both the subject and background are dark in real life, in this hypothetical example.
So, the photographer has to do more than simply ask him- or herself whether a histogram shows a preponderance of middle vs. light or dark tones. He or she needs to think of the subject and scene, and whether a properly rendered image of that scene would tend to have a lot of dark or light-toned values. If the subject at hand is that white rose against a white background, and a histogram shows the bulk of pixels in the middle of the image, it's a clear sign to the photographer that this particular image is under-exposed–instead of mostly light-toned pixels, the actual image file consists of mostly middle gray pixels.
How to read a histogram: off-scale brightness areas
One of the easiest ways to quickly use a histogram is to look at its graphic values, and simply ask if any of the bright areas on the right side, or dark areas on the left side, are butting-up against the far edges of the histogram itself. Either condition indicates SOME area of the scene is reproducing as an over-exposed white, or an under-exposed black tone.
Now, it's again up to the photographer to think for a moment. If the scene is a snow-covered field or a bride in a white wedding dress, and the bulk of the histogram is flush against the right edge of the graph and visibly cut-off along the right, it's a clear indication that the image is over-exposed, and that another picture should be taken with reduced exposure.
But, it's entirely possible to have a properly exposed scene with bright objects, such as reflections off of shiny objects, or even the sun in the frame, which do appear as over-exposed areas. If these are the only areas in the scene that are washed-out, the histogram may be telling you the exposure is actually OK! So don't immediately assume that there's a problem if part of the histogram is cut-off along the left or right side, until you've thought about the scene and subjects within it.
How to read a histogram: overall brightness range
A histogram is a great tool for judging whether the brightness range of a scene will fit within the dynamic range of the camera. If the exposure graph fits within the histogram's left and right margins, most likely we have an easily workable and printable image. If the histogram is pushed up against either side, some parts of the image will be too light (right side) or too dark to reproduce with detail.
A "good" histogram doesn't necessarily have to stretch entirely from the left to right side to indicate proper overall exposure. But if its peaks, large and small, all fit within the left/right borders, it's telling you that nothing in the scene will reproduce as a washed-out white, or a black with zero detail.
Histograms – brightness only, or color information as well
Conventional histograms measure only brightness, as recorded (or about to be recorded) in an image file. But, it's possible to view histograms that not only show overall brightness, but the brightness of each color channel. Canon EOS cameras offer the choice to display a brightness histogram or a color (RGB) histogram.
For most users, the Brightness histogram is sufficient to evaluate exposure when reviewing images. But, there are instances where it can be useful to view a graph of all three color channels. The RGB histogram can inform a user when there's a color shift (preponderance of one color vs. others), which is useful when trying to white-balance a known neutral-colored test subject. It can also be helpful for determining whether any single color channel is at risk of being rendered too light or dark.
Color histograms can potentially be nearly identical for two images containing different subject matter if they share very similar colors, recorded at similar exposure levels. Also, similar objects with different colors may be indistinguishable in color histogram comparisons. Based solely on a histogram, it may not be possible to distinguish a green apple on a red plate or a red apple on a green plate.
Remember, while there are times that you'll prefer to shoot JPEGs, RAW files will always offer more exposure latitude and more room to correct potential problems. Correcting improper exposures during image editing in the computer has limitations and often results in a loss of image quality. Histograms give the photographer a tool to evaluate and perform corrections immediately after an image is taken, which means that less time is later spent at the computer trying to optimize images.