There are Two Types of Histogram

Definitely be aware of this - One type of very common histogram does NOT show your color image's actual real data.

Histograms can show just one of the individual Red or Green or Blue channels, but they typically show a three channel combined value.

There are two types of these combined histograms, not at all the same thing:

There is a "RGB" type (second below, Adobe is a popular example) where the three RGB channels are shown superimposed on the histogram (overlaid in place, actual real data, but gamma encoded), so that any red pixels are still shown at the red pixel values, etc. The colors may not be distinguishable there as such, but they do not shift position. This is the useful type.

And there is the very common Grayscale or Luminosity type (two names for same thing, next below), where the RGB channels are combined by first converting to compute luminosity. Luminosity is a measure of relative brightness (how brightness appears to us now). In the case of image data, it computes simulated theoretical grayscale data using the NTSC television standard formula, to create the new simulated grayscale pixels:

Luminosity = 0.3 R + 0.59 G + 0.11 B

This formula was developed by television as the accurate way to convert RGB to grayscale, according to luminosity. This is the way the RGB color components of every RGB pixel are "weighted" to create the pixel's one gray value. That luminosity value result represents the relative brightness of gray that the colors would appear to the human eye, based on our eye being more sensitive to green, and less sensitive to blue. In photography, luminosity is about how colored objects appear on B&W film. For example, the relative brightness of red lipstick, green grass, or blue skies, etc. - could be made to come out any dark or light value, but there is one correct way our brains perceive the brightness of that actual subject. Luminosity computes these colors to come out the same brightness as grayscale does, matching their brightness the way it is perceived by the human eye.

Example: a Nikon D300 camera rear LCD display showing result of a picture entirely of this one red color for simplicity.
This RGB color is (255, 103, 95). Or more specifically, the normal exposure of this picture was such that red came out clipped at 255.

The D300 camera LCD display can show four histograms as above, which shows the three individual RGB channels (255, 103, 95), plus it also shows the computed Luminosity composite (on left side - the theoretical grayscale values of this image). No bets about every model, but the Canon cameras appear to do the same thing.

Note that when the Nikon camera is showing only the single graph, it is luminosity. The screen above may be named RGB (just meaning it also shows the three RGB channels, but individually, not combined). It also still shows Luminosity there, which is not RGB data - it is the grayscale equivalent. The formula computes this actual RGB color (255, 103, 95) to show as

Luminosity = 255 x 0.3 + 103 x 0.59 + 95 x 0.11 = 148   (one grayscale luminosity value).

This indicates the relative brightness of this color is slightly above midrange - the relative brightness of this color, as perceived by the human eye and brain, and on B&W film (negatives are inverted). But the actual RGB data is something rather different.

Note this Luminosity display does not give any hint that one of the channels is clipping. We can select One channel to watch here. The scene was made color red here, because red flowers are often tough this way. Don't use the camera's single histogram to determine clipping. In the camera, you definitely want to watch the three RGB displays instead.

The Luminosity value shows 148 above, using the standard formula just shown. The luminosity 148 sum is not any one of the RGB values - it is a computed theoretical value (it is the grayscale equivalent value). The number 148 certainly looks safe, not far above the middle, but nevertheless, the Red channel is obviously clipping. The Red 255 value contributes to the gray sum only as its 0.3R component, and green and blue components are also added in. FWIW, if you are photographing a neutral gray card, all the RGB channels and composite should show the same value, equal RGB components, if it is neutral gray (no color cast). So the three numeric coefficients of luminosity add to 1.0, so the math can only work out to be the same value of gray. But colors make it be rather different.

Most normal images have wider content and a smoother response and may appear "different" than these three spikes, but any image is the same concept. This new luminance sum shows the "relative brightness", specifically, the brightness that color would appear in B&W film. Here its luminosity peak is near the middle - and the red display is in fact not bright and not dim. It may be a theoretical concept, and the luminosity formula and histogram has use to show the perceived brightness, but the physical reality that exists is the three RGB channels, and RGB and clipping is the important factor in the camera today.

Note in the Nikon camera, you can zoom in on the image, and move the display around, and then the RGB histogram represents only the part you are viewing then. Adobe software is similar, but it is not about zoom, instead it (Levels and actual Histogram menu) shows the histogram for the selected area.

Second Type of Histogram - Useful for our photography purposes

Adobe programs (and several others) instead show the RGB histogram type, where the three RGB channels are overlaid in place (no computations). This same orange image above looks this way in ACR (Adobe Camera Raw, later, when out of the camera):

Adobe Camera Raw

And Photoshop Levels is also RGB type, same thing, same data, we still see the real data, and the real clipping - it just does not indicate the colors (below). The same overlaid data is shown as one gray curve, but it is not the grayscale luminosity curve (unless it was a grayscale image).

Adobe Photoshop CS5 Levels

Both are showing the same image, and both show our data is clipping (because the RGB channels are simply overlaid in place, so red is shown where red actually is). In contrast to the Nikon camera's single histogram showing luminosity (repeated at right), is a computed theoretical brightness value (grayscale), which does not show real RGB data or the clipping.

Repeating so you don't miss it, this type above shows the three RGB channels overlaid in place, and in their correct positions. We may not know or care which one is red, but we can see the clipping at the 255 end - which is pretty much our only reason for looking. So be aware that most cameras also show the individual RGB channels too, and that only the individual RGB histograms show the actual clipping information we seek. The camera's single display is luminosity, which may not show clipping.

Or, the Photoshop "Window - Histogram" menu has an option to show luminosity, but it is not the default for RGB images (grayscale data is necessarily luminosity). The free editor program Faststone also has options to show histograms either way, but its Levels tool shows RGB. The situation is improving slowly, but Luminosity is still the default view for histograms in relatively many lesser editor programs, and cameras, and scanners (just how it was always done).

Again, the big deal is that this RGB histogram clearly shows the red channel is clipping, but the Luminosity histogram view in the camera gives no clue of this. Still good to know it though.

These RGB and Luminosity histogram modes are easy to detect, if you can inspect to see if the composite graph contains peaks corresponding to the individual channels, or not. But anything neutral color, or with a lot of white, will be very confusing to see this - white is equal RGB components, effect will be much less apparent, might appear OK when it is not. But if you create a frame of pure red (255, 0, 0), if which shows on histogram near 0.3 x 255 = 77 instead of 255, that is Luminosity. Photograph that color on the monitor screen (carefully overexposing just a tiny bit) to see how your camera histogram reacts to it.

Or if more convenient, here is that same actual picture above from the D300, already prepared if you want to look at its histogram other ways. I measure RGB as about (255, 75, 50), depending on where you click, but as above. Displayed small here, but if you save it (Right click it, Save Image As), it is a 1280x850 pixel image, 70 KB. It is already overexposed a bit, so there may be slight histogram changes - which don't matter. But if you show it large on the computer monitor, and photograph that red screen with your camera, carefully overexposing this second time a bit again, you can see how your camera histogram reacts to it. Be sure to look at the RGB channels too.

This is the PhotoMe browser histogram, clicking on the red image here on this web page. The weak blue and green components keep it from being a pure red. But it is NOT showing one luminosity peak near mid scale, so we know the type of histogram. In the camera, a closeup photo of a manila envelope or brown corrugated cardboard box is similar, apparently a single color, but three, or at least two distinct RGB peak values, as opposed to one luminosity color.

It is important to understand the histogram view in the camera, so that we can use it to control exposure, to avoid clipped channels. Be aware that Luminosity histograms may not show that. Photographers simply don't much care about the rest of the histogram data curve, which will vary greatly in different images, but clipping is important to us. The luminosity curve won't show clipping (not meaningfully).

BUT ... when concerned about excessive exposure clipping of the bright tones in our digital camera images, the Luminosity histogram is not the right choice. It may show the way our eye perceives the brightness of the tones, but it does not show actual the RGB values stored in the data, and may not show any clipping. You want to pay attention to the RGB values, if concerned with clipping.

Regardless of histogram type, all data in all RGB images is gamma encoded.

So there are two ironies about histograms ...

1) Our only use for histogram in the camera is to judge clipping (or proximity to clipping), but the standard luminosity histogram cannot show that for RGB data.

2) The RGB data is gamma encoded, so that middle range is not at middle, but instead up closer to 3/4 scale.

It all goes better when we understand what we see.