Histograms – A page about the interpretation of histograms in digital photography.
A histogram, as used in digital photography, is a tool for evaluating exposure and visualizing the distribution of tones in a photograph. Tones range from black point to white point along a continuum of almost countless degrees of brightness within an image. However, for the sake of efficiency, histograms often display brightness within a span of 256 variants. Additionally, and for even greater simplicity, photographers often refer to tones as being within commonly recognized tonal categories. These are: black point, shadows, threequarter-tones, midtones, quartertones, highlights, and white point. After a time, we will find ourselves using the these terms when discussing a photograph. For example, a professional photographer may say, “There is too little detail in the shadows of this image.” “The histogram reveals that your exposure is way off, highlights are clipped past white point!” Or, “If we increase brightness in midtones using Photoshop’s Curves adjustment layer, the image will look more pleasing.” Could we have a scene with no black point, no white point, or neither? Or, for that matter, no or very few midtones? Yes. As with any craft, we must achieve a level of understanding and experience to use tools effectively.
A histogram is NOT an image of your image, nor is it useful for evaluating the merits of an image. But it is informative concerning the exposure settings you have chosen in a camera or when altering the brightness (up or down) of tones in a photograph during post-production processing. After some study and experience, when referencing a camera’s histogram in the viewfinder, we may say to ourselves, “I don’t like this.” “This is as good as I am going to get.” Or “This is sweet!” Unfortunately for Raw shooters, digital camera histograms are based not upon the Raw file but upon a JPEG “thumbnail” of the Raw file. It is only in post-production that we can see a histogram of our Raw file.
I have to keep reminding myself that a histogram is not about color or location within an image. It is about the brightness of individual pixels. A pixel (short for picture element) that is bright enough to be in the highlight category, for example, may come from any area in the image. Even in a histogram that displays the three (red, green, and blue) color channels of the RGB color model (used by almost all photographers), it’s all about brightness within those color channels. A burgundy pixel (R: 128, G: 0, B: 32 by the way), which is clearly not as bright as red (R: 255, G: 0, B: 0) is displayed in a histogram according to its level of brightness not because it reminds us of Cabernet Sauvignon. Recall, that in computer speak, 0 is a number so there are 256 numbers between 0 and 255.
Imagine the histogram as a digital post office. There are 256 possible tone address boxes at this post office. Post box 0 (Black Point) is pure black. As black as the bottom of a coal mine. Post box number 128 (Midtone) is exactly halfway… 50% brightness! And post box 255 (White Point) is white as the driven snow. Now when an image arrives at this post office, the poor overworked clerk must study every pixel in the image then drop it into one of 256 available boxes. The more pixels there are in any one tone or tonal area, the higher the stack of mail becomes in that area. In other words, the height of any tonal area in a histogram reveals the number of pixels of that tone within the image. This is why, if we create a blank (all white) new file in Photoshop, the histogram displays only one thin line at the extreme right…white point. In theory, we could follow that thin, one pixel line, up into outer space. But Photoshop’s histogram height is relative, so pixel count does not go through the “roof” of the histogram, Other post-production programs allow pixels to vector upward past the visible boundary of the histogram display. In a 75-megapixel image, the clerk would spend an entire lifetime evaluating each pixel for brightness!
Fortunately for humans, our camera and all post-production software programs, create a histogram for us instantly…on the fly as they say. But we ignore it. Why? We ignore the histogram when using our camera or when we tinker with the image in a post-production computer program because we really don’t understand it and we don’t realize how very important it is.
Observe the luminosity histogram in the upper right corner of the image. It tells us that the image has most of its tones (pixels) in the threequarter-tone, midtone, and quartertone areas diminishing at both ends toward black point on the left and white point on the right. There are no clipped shadows or clipped highlights as no pixels are beyond black or beyond white. I took the above image in a grassy meadow in the Santa Monica Mountains north of Malibu, California. It is a very telling image because the associated histogram screams Perfect Exposure. But how could I go wrong? A digital camera’s light meter is tuned to about 12.5% reflectance (anticipating an average amount of light reflected toward the camera) and a grassy meadow has a reflectance of about 18%. But Nikon’s matrix metering would probably rate this scene a bit closer to 12%. In sum, there were no exposure challenges. The scene has a relatively small black point and a modest white point as well. Not that black and white are undesirable in an image but, rather, too much would obscure the meadow. In fact, too much in any typical photograph would be counterproductive unless there is a very dark or very bright subject of interest to be revealed. Yet almost every image benefits from some black and some white to give viewers a frame of reference.
Not long ago, a fellow photographer criticized the above “normal distribution” of tones. “There is no such image!” he said barely containing his rage. “And it is pointless! Every image is different and has a different distribution of tones!! Shame on you, shame, shame!” Of course, I flogged myself. Yet my friend misunderstood. A normal curve is a statistical instrument not a recipe. If we averaged five-hundred photographs taken by professionals, I suspect the normal distribution of tones in those photographs would look something like our curve.
We can learn a thing or two from this normal curve even if it can only be accurately represented in real life by a grassy meadow at noon. Think of this normal curve as a lump of clay. We could mold it toward Black Point (underexposure) or White Point (overexposure) effortlessly when using our camera in Manual mode (or by selecting +EV or -EV Exposure Compensation in other modes such as Aperture Priority). Obviously, this would change the distribution of tones. Most often, in real life, our lump of clay would be much more complex. Because, alas, life is complex. But the lump of clay reminds us of one thing regardless of circumstances. If the histogram has a flat line at either end we have lost some dynamic range (no information) and this must be addressed either by taking another shot, having corrected our exposure, or in post-production using the host of tools provided to alter tonal distributions. Let’s look closely at some histograms….
In the above image, some rascals have pushed our lump of clay down in the middle right over midtones! Those rascals are both ambient light at the scene and my post-production conniving. The reflected light in this scene had a respectable but less than abundant supply of midtones. But look how dramatic the image is! You see, there is no good or bad in tonal distribution. Existing tones may be either useful for communicating our subject of interest or not useful for communicating our subject of interest. It’s a judgement call. I could have brightened shadows in this image more than I did in post-production, but I like the drama. The fade to black. Your call!
If you look casually at the above histogram. you may conclude that highlights are clipped past white point, but this is not the case. Highlights are not quite past white point. Many photographers endorse “pushing highlights” or “exposing to the right” (ETTR) as three-quarters of our tonal data is gathered from midtones to white point, while that darker, noise gathering half from black point to midtones holds only one-quarter of an image’s data. It’s this simple, more light results in more useful data, less light results in less useful data. The “signal to noise ratio” diminishes in a poorly illuminated scene and an image will contain more noise. We may increase ISO, but past some point image quality suffers as noise dominates the image. In brief, I do monitor the histogram in my camera’s viewfinder and often but not always nudge highlights closer to the white point.
Oh no! Here is that awful histogram curve again…. Well, OK, let’s make the best of it. Often, a camera’s luminosity histogram (gray thing, at middle gray by the way) is not visible in a camera’s viewfinder by default. Marketing doesn’t want to frighten away timid customers with all that techy stuff. Once you have found it in the camera’s menu and turned it on you are in business. In the eyepiece or LCD the histogram display is somewhat small so you may not see it at first. My first thought here is to have you mind the right side of your exposure, as revealed in the histogram, so there is no significant gap between the right side of the histogram and the right side of the histogram box (white point)…as illustrated above. This exposure is shifted toward underexposure.
To push the histogram to the right we must brighten (open up) our exposure. In Manual (M) mode this is an easy fix. We either select a slower exposure time (sub-command dial on a Nikon) or a larger aperture (main command dial in front). Recall a larger aperture is represented by a smaller f-number to keep us on our toes! In all other modes we can use that shy little Exposure Compensation button on our camera (typically) near the shutter button where all the other important little buttons are located. Push the button while turning the command dial (Nikon) and Presto! You can move that histogram as easily as the repo man moves your BMW. A -EV (Exposure Value) moves the exposure toward black point and a +EV moves the exposure toward white point. In many exposures we may choose to move the histogram so that it nudges but does not knock over the right side of the histogram box. Now you are exposing to the right. But, what if there is no white point in our scene? What if there are few highlights? In truth, the histogram is a tool and, like any tool, there must be a brain behind it. When we use a hammer, we don’t hit every nail as hard. So evaluating exposure while using the histogram as part of that process will take some practice. The ISO button can also help here if needs-be. Recall your exposure triangle?
Oh, by the way. You may want or not want to keep the exposure compensation you have chosen. It may be that your next shot will also benefit from the exposure setting you have chosen. Exposure compensation is also used to override the camera’s meter under exposure conditions that mess with its little brain. While shooting downhill skiers, for example, you will need to use, perhaps, a +2EV or +3EV (exposure compensation +2 or +3 Exposure Value) because a camera’s metering system may well be fooled by the highly reflective snow in such a scene and underexpose your image. Conversely, a steam locomotive absorbs light so a -3 EV may be appropriate.
In the above photograph, our lump of clay is pushed toward black point a bit more than I like. Enough to deepen highlights to the point that I want to brighten the image from midtones to white point using a Curves adjustment layer. Curves, as a post-production tool is ubiquitous these days in post-production software. Say, what is the subject of interest here anyway? That stick thing? Oh, ok.
We are viewing both “luminosity” and color histograms on this page. Red, green, and blue light have different reflection characteristics and a luminosity histogram compensates for this by providing a histogram that displays tones as we humans perceive brightness. Human perception of the brightness (luminosity) of a color can be described by a ratio of approximately, R: 2.1, G: 7.2, B: 0.7 (total 1). We humans see green tones very well. You would think we evolved in trees! There are twice as many green photosites in your camera’s sensor than red or blue (Bayer mosaic).
We can always see individual color channel (RGB) histograms in our camera or in post-production. More than a few photographers suggest we are less than responsible not to do so. Below a color histogram.
Above we see a RGB color channel histogram. A tad more red for obvious reasons. The exclamation symbol at the upper right of the histogram reminds us to “refresh” the histogram as the image has been changed. I added some contrast using a Curves adjustment layer and the classic “S” curve (see lower right).
Above we have a masterful work of modern art! The histogram function does not “know” a graphic image from a photograph. Both can be saved as a JPEG or .jpg image file. We see from our Levels adjustment layer that the image has a black point and white point. Where colors cross paths, they generate tonal variations as shown in the color histogram. But overall tonal variation is quite limited as we might expect given that every color is very bright. Only a few post boxes are used.
This graphic includes the three colors of the RGB (additive) color model (red, green, blue) and four colors of the CMYK (subtractive) color model (cyan, magenta, yellow, and black). In post-production, black is (R: 0, G: 0, B: 0) and white is (R: 255, G: 255, B: 255). I made the colors as bright as possible. For example, Green is (R: 0, G: 255, B: 0). Magenta is (R: 255, G: 0, B: 255). Of course, no four color press is going to print at that level of brightness. Magenta will print at approximately (C: 12, M: 79, Y: 0, B: 0) using my four-color inkjet printer. The subtractive process will create the perception of greater brightness and color saturation in a print, within limits. The histogram for this image is the best Photoshop can do to represent the tones present.
A photograph with only seven colors would disappoint most fashion models. Recall that Windows 3 allowed 256 colors for an image or 8-bit color depth (2x2x2x2x2x2x2x2 = 256) for all RGB color channels: (RRR,GGG,BB). For example, 0110000 is code for red in 8-bits for you soon to be Geeks. Today, an image may have 16-bit depth PER COLOR CHANEL for a total of 281 trillion possible colors.
With solid colors, having no anti-aliasing and no overlapping, our luminosity histogram has little tonal data to report. But the proportional illustration of pixels is to be expected.
Now here we have something! A deep blue sky and a very bright sunlit building. The tonal peaks over 3/4 tones and 1/4 tones seem distracting to me. Midtones could be better represented. A different time of day or a different location may be beneficial. Perhaps during the “Golden Hour” for photographers, the last hour before sunset and the first hour after sunrise! Note how the sky transitions from cyan at the horizon to blue above. To render a darkened sky in black and white photography, we reduce both cyan and blue color channels.
The Above Levels adjustment layer from Photoshop CC shows us that our next image has no black point or white point and that shadows and highlights are drastically underrepresented. The more flat line (or worse, no line) we have, the more we will sacrifice tonal information to sort it all out in post-production. But there is still quite a lot of information here.
In the above image, as shot, the lighting was poor and the exposure unimaginative. Tonal information on either side of midtones suffered. It would have been wise to push this exposure to the right given the prevailing light. However, I will “remap” the black point and white point using Levels adjustment layer’s sliders. Then apply a Curves adjustment layer for contrast and brightening. See my mediocre result below…
The flat lines at either side of the tonal range in our image DO contain some tonal information. The empty spaces do not. It is not necessary to sacrifice this information when correcting black and white points. Tonal information in the above image did suffer somewhat due to the initial exposure and my brutish editing technique. Was it really necessary to use the Levels sliders so aggressively? No. In truth, we can use Levels sliders or Levels or Curves “eyedroppers” to set a new black point and white point. While the use of the eyedroppers requires some practice, the results are often better.
However, if we use an eyedropper to set black point and white point, we are also color correcting the image simultaneously. Be careful what you ask for. If your white point has a touch of pink, the white eyedropper will color shift the entire image to match your adjustment request which is to make that pink cast go away. This can be a very effective way to experiment with different “looks” for an image. Often, when using a Curves white eyedropper over an area or spot of white, the entire image seems to pop into a much better color presence. Occasionally, I end up with three Curves adjustment layers when using Photoshop, one to set white point using an eyedropper, another to set black point using an eyedropper, and a third to adjust contrast using the classic “S” shaped curve. Creating contrast in an image during post-production always discards some tonal data.
Can’t find a candidate point for applying the Curves black point (or white point) eyedroppers? Use a Threshold adjustment layer and mark the spot using Guides as cross-hairs. How do you find guides? Drag them from rulers. Don’t use rulers? Say what! View > Rulers…seriously.
The above image is, generally, shifted toward underexposure, as is illustrated in the histogram, but the brighter areas capture our imagination. I like it the way it is. The histogram also shows us that, although a lot of blacks and shadows are clipped, areas of brighter tones are sufficient to please the eye.
Black and white images have a histogram too! Here we see that some shadows and highlights are clipped. If we aggressively alter an image’s tones, we sacrifice tonal transitions and head toward “posterization,” where solid colors begin to replace “continuous tones” and detail. But no matter, high contrast is the dominant theme in the above photograph. Not every photograph is so lucky.
I used a Curves adjustment layer to add extreme contrast to this photograph. It is beginning to posterize as my efforts have stretched tonal information throughout the image. If you look carefully at the histogram, you can see the gaps left by my handiwork. These gaps reveal missing tones. As tones are information, these gaps actually betray missing information…photographic information. The message is…perform adjustments with an eye toward what is being gained vs. what is being lost. Our great achievement in one aspect of an image may have us indited for crimes against photography in another. This is also true for color adjustments and use of color saturation. If we are not cautious, we can create a color Godzilla on our computer screen!
Metering modes are a must know for photographers!
Below, the exposure compensation function in our camera can be used to aid exposure.
Above for reference.
Photography can be fun, profitable, creative, and technically superior at the same time. Contact me for a custom one-to-one camera class in Bakersfield, California. Please text me at: 661-303-9210 (preferred) or email me using: email@example.com for an appointment.
Back to the home page
Back to top of page
All images and illustrations are by Ed Ruth…for better or worse!!