A histogram is a graph of the brightness in a photo: dark tones on the left, light tones on the right, and the height at each point showing how many pixels have that brightness. It matters because it tells you, in a second and in any light, whether your exposure has kept the detail you care about, which the screen on the back of the camera cannot do reliably.
The snow scene above has almost all of its graph piled on the right, and that is exactly right, because the scene is almost all white. Reading a histogram is never about finding one correct shape. It is about asking two questions: does the shape match the scene, and is anything important cut off at either edge?
How to read the two axes of a histogram
The horizontal axis runs from pure black at the left edge to pure white at the right edge. As a rough guide, the left third is shadows, the middle third is midtones and the right third is highlights. The vertical axis is a count of pixels, so a tall peak means a lot of the picture has that brightness and a flat stretch means almost none of it does.
- Piled at the left. Most of the picture is dark. That is correct for a night street or a cave, and a sign of underexposure for a daylight portrait.
- Piled at the right. Most of the picture is bright. That is correct for snow or a white wall, and a sign of overexposure for a moody evening scene.
- A hump in the middle with nothing at either end. Most tones are midtones and the scene has little contrast. This is typical of fog, haze and flat overcast light.
- Two humps with a dip between. A bright part and a dark part with few midtones, such as a bright sky over a dark skyline or a window in a dim room.
The vertical axis has no numbers on purpose. Each histogram is scaled so that its own tallest peak reaches the top of the box. A peak that looks tall in one photo and short in another tells you nothing about which photo has more of that tone; it only tells you how that tone compares with the rest of the same picture.
The horizontal axis is not divided into stops
Each gridline on the graph is not one stop of light. The graph shows the brightness values of the processed picture, and those values are gamma encoded: the dark end is stretched out and the bright end is squeezed together. The table shows where tones land when they are a set number of stops below pure white, using the standard sRGB curve with no extra contrast added.
| Stops below pure white | Value out of 255 | Distance across the graph |
|---|---|---|
| 0 | 255 | 100% |
| 1 | 188 | 74% |
| 2 | 137 | 54% |
| About 2.5 (middle gray, 18% reflectance) | 118 | 46% |
| 3 | 99 | 39% |
| 4 | 71 | 28% |
| 5 | 49 | 19% |
| 6 | 34 | 13% |
| 8 | 13 | 5% |
Two things follow. The top stop of brightness takes up the whole right quarter of the graph, so a hump that sits three quarters of the way across is already only one stop from clipping. And a middle gray subject lands a little left of center, not dead center. Real cameras add their own contrast curve, so the positions move a little, but the pattern holds: one stop is a long way at the right of the graph and a short way at the left.
It counts tones, not places
A histogram knows how many pixels have each brightness, and nothing about where they are. If you shuffled every pixel of a photo into random positions, the result would look like static and its histogram would be identical to the original. So a hump on the right does not tell you whether the bright tones are in the sky, on a face or on a white shirt. Your eyes and the picture tell you that. The histogram tells you how much, and whether any of it is lost.
Matching the shape to the scene: high key, low key and flat light
Snow or a white studio background should have its histogram on the right. A night street or a moody still life should have most of its pixels on the left. Dragging either toward the middle makes the picture duller and destroys the mood.
The catch is that a camera meter aims to make whatever it reads come out as an average gray. It pulls snow down toward gray and lifts a dark scene up toward gray. The histogram is how you catch it. If a snow scene shows a hump in the middle and nothing near the right edge, the snow has been recorded as gray and you need to add exposure. If a night scene fills the middle of the graph, night has turned into dusk and you need to take exposure away. Exposure compensation is the quickest way to apply the correction in an automatic mode, and the page on metering explains why the meter behaves this way.
So a tidy, centered graph can still be the wrong exposure, and a graph piled at the left can be exactly right. The test is always the scene, not the shape.
Try it in the camera simulator: the meter turns this snow gray. Switch on the histogram, then use exposure compensation until the snow is white and the graph leans right without being cut off at the wall.
Width tells you about contrast. A narrow graph that stays away from both walls means the brightest tone in the scene is not far from the darkest, as in fog, mist or a flat overcast day. A graph that spreads from wall to wall means deep shadows and bright highlights in the same frame. Fog should look soft, and a backlit scene is full of contrast. A wide tonal range gives a picture depth, but the pixels need not be spread evenly across it.
Want a bright or dark result on purpose? The page on high key and low key photography goes deeper into both, from lighting to editing.
Clipping and blinkies: when the graph hits the wall
Clipping happens when a tone is brighter or darker than the camera can record. Everything past the limit is stored as the same value, pure white at the right or pure black at the left. On the histogram it shows as a tall bar jammed against the edge, or a graph that is still climbing when the wall cuts it off instead of fading to nothing before it.
A clipped area has no texture, because every pixel in it has the same value. A white dress that clips becomes a flat patch, and a sky that clips becomes a blank hole. Clipped shadows are just as lost: brightening a pure black area in editing gives you a flat gray, not detail. The difference is what is next to the wall. Shadows that are dark but not clipped can usually be lifted in a raw file, at a cost in noise. Highlights that have clipped in the raw file cannot be rebuilt at all, and pulling down near clipped highlights often shifts their color. So when you cannot keep both ends, protect the highlights that matter first. The page on blown highlights shows how to spot and avoid them.
Clipping is not always a mistake. In the skyline above, the buildings are a pure black silhouette, so the histogram has a tall spike at the left wall, and that spike is the whole point of the picture. A glint on chrome, the sun on water or a lamp in a night scene will clip in almost any exposure too. Chasing them would darken everything else. These small bright points are called specular highlights, and losing them costs nothing.
Most cameras can flash clipped areas on playback. These highlight warnings, or blinkies, answer a different question from the histogram. The histogram says how much of the picture is clipped. The blinkies say where. Use both: if the blinking area is a lamp or the sun, accept it; if it is a face or a white shirt, take some exposure away. Many mirrorless cameras also offer zebra stripes, which mark bright areas before you press the shutter.
The practice game below marks an edge in red when more than half of one percent of the pixels are clipped there. That is a useful line for a first glance, not a law. A few clipped pixels in a glint are harmless, and the same count across a cheek is a problem.
Luminance and RGB histograms: how a color can clip while the graph looks fine
The default histogram on most cameras is a luminance histogram. It combines the three color channels into one brightness value for each pixel, and the channels are not counted equally: in the common video and sRGB standard, green counts for about 72 percent, red about 21 percent and blue about 7 percent. That matches how bright each color looks to the eye, and it makes the luminance graph a good summary of overall brightness.
It also means a single channel can clip while the luminance graph stays clear of the wall. An RGB histogram shows the red, green and blue channels separately and catches what the luminance view hides. A red flower, an orange sunset or a neon sign can push the red channel against the right wall, and the clipped color shows as a flat patch with no shading, as if painted on.
Blue does it too. In the maple above, the pale sky has clipped in the blue channel across about a quarter of the frame, yet the luminance graph shows almost no clipping, because blue counts for so little of luminance. When one channel’s curve climbs the right wall, lower the exposure until it pulls away, or accept the clip if the area is small. Check the RGB view whenever the scene has strong, saturated color.
Practice: read the histogram
Ten rounds of three kinds. In some you pick the photo that made a histogram, in some you pick the histogram that belongs to a photo, and in some you say what a histogram is telling you. Each photo is shown from two stops darker to two stops brighter, a red edge marks clipping, and every answer comes with an explanation.
Did a red edge catch you out? Clipping explains what it means when the graph is cut off at either wall and how to avoid it.
Want to move the exposure yourself and watch the graph follow? The camera simulator has a histogram you can switch on while you change the settings.
Where the camera histogram comes from, and why it is cautious for raw
The graph on the back of your camera is not computed from the raw sensor data. It is built from the processed preview, the same small JPEG you see on the screen. That preview has white balance, contrast, saturation and sharpening applied, using whichever picture style is selected.
That has two effects. A vivid or high contrast picture style makes the histogram look more clipped than the data underneath it. And a raw file usually holds some headroom beyond what the preview shows, so the histogram can warn about clipping that is not in the raw file. In practice the camera histogram is conservative for raw shooters: if it just touches the right wall, the raw file may still hold detail there. How much headroom depends on the camera, so the only way to know yours is to shoot a test series of a bright subject, open the files in your raw editor and see where detail really runs out.
If you shoot JPEG, the histogram is an accurate description of the file you will get. If you shoot raw, choosing a neutral, low contrast picture style brings the graph closer to what the raw file holds. Some go further with a custom white balance called UniWB, which turns previews green but brings the channel graphs closer to the raw data. The page on why raw files look different from the camera preview explains the gap in more detail, and raw versus JPEG covers the other differences between the formats.
On many cameras, zooming into part of the picture during playback makes the histogram show just that part, so you can check a face or a patch of sky on its own.
Expose to the right: using the histogram for cleaner shadows
Noise in a photo comes mostly from too little light. Light arrives in random packets, and the fewer the sensor collects, the larger that randomness is compared with the picture itself. So the more light you give the sensor without clipping what matters, the cleaner the shadows will be once you bring the picture back to the brightness you want. Exposing to the right means giving the scene as much exposure as the highlights allow, with the histogram close to the right wall but not climbing it, then darkening the file when you process it.
The table above shows why this takes care. The brightest stop fills the right quarter of the graph, so the last stop before clipping is a short distance in light but a long distance across the histogram. Raise the exposure with a slower shutter speed or a wider aperture, check the graph, and stop when it nearly reaches the right edge. Raising the ISO is not the same thing: it makes the image brighter without gathering any more light, and it brings the clipping point closer. Work in raw: a JPEG made this way looks washed out and has little room to come back down.
It suits still scenes with time to check, such as a landscape on a tripod, and is a poor fit for fast action. The full technique is on the expose to the right page.
When the scene is wider than the camera can record
Sometimes the graph hits both walls at once, and no exposure setting moves it clear. Then the scene has a wider range of brightness than the sensor can record, its dynamic range. Moving the exposure only slides the problem from one end to the other. You have to decide what to give up, usually the deepest shadows, or change the light with fill or a graduated filter, or take a bracketed series of exposures and combine them later with exposure blending.
Slide the exposure below and watch the histogram: with a scene wider than the sensor, moving the exposure only trades clipped highlights for blocked shadows, until you bracket and merge.
Using the histogram while you shoot
- Take a test shot at your best guess for the exposure.
- Look at the histogram, not the picture. A screen looks brighter in a dim room and darker in bright sun, but the graph does not change with the room.
- Check the right wall first. If the graph is cut off there and the clipped area matters, reduce the exposure.
- Check the left wall. If it is cut off there and the highlights have room, add some exposure.
- Ask whether the bulk of the graph matches the scene: right for snow, left for night, spread out for a sunny day.
- Adjust with shutter speed, aperture or ISO, or with exposure compensation if your mode sets the exposure for you, and check again. Repeat when the light changes.
This takes seconds once it is a habit, and beats squinting at the screen, which people call chimping. A mirrorless camera, or a DSLR in live view, can show the histogram before you press the shutter, so you can set the exposure without a test shot. Through a DSLR’s optical viewfinder you only see it afterwards.
You do not need to check every frame. In even light the meter does well. Check when the light is tricky: snow, sand, night, backlight, a subject much brighter or darker than its background, or an exposure too long to judge by eye, like the 3 second street scene above.
Is the camera reading the wrong part of the frame? The way it measures the scene decides where the graph lands, and the page on metering modes explains how each mode reads it.
Using the histogram when you edit
In many editing programs the histogram is also a control: hovering over a region shows which slider moves those tones, and some let you drag on the graph itself. Turn on the clipping overlays and lost highlights and shadows show in color on the picture.
The classic levels control has three triangles under the graph. The black triangle sets the darkest tone, the white triangle sets the brightest, and the gray one moves the midtones lighter or darker without touching the ends. Dragging the black and white triangles in to meet the edges of the data stretches a flat picture to the full range. The page on the black point explains how the end points work, curves and levels covers both tools, and the step by step version is in setting your black and white points.
Heavy editing leaves marks on the graph. A comb of thin gaps and spikes means the tones have been stretched too far, and smooth gradients such as a sky may show visible steps called banding. Smaller changes, or working from a raw file with more bit depth, reduce it.
Common mistakes with the histogram
- Trying to make every histogram a centered bell curve. The right shape depends on the scene. Snow should lean right and a night street left. A centered snow scene is gray snow.
- Trusting the screen over the graph. The screen changes with the room and with your eyes. The graph does not.
- Panicking over a few clipped pixels. A glint, a lamp or the sun is fine to lose. Ask whether the clipped area is something you care about, and use the blinkies to see where it is.
- Thinking a clear right wall means nothing is clipped. One color channel can clip while the luminance graph looks safe. Check the RGB view for saturated color.
- Reading the axis as stops. The graph is gamma encoded. One stop below white is already a quarter of the way in from the right edge.
- Comparing peak heights between photos. Each graph is scaled to its own tallest peak, so heights only compare tones within one picture.
- Expecting it to tell you about focus or color. The histogram knows nothing about sharpness, composition or whether the white balance is right. Zoom into the picture to check focus.
Try this: make the meter get white and black wrong
You need a sheet of white paper, something black and matte such as a dark sweater, and even light. A phone works if its camera app shows a histogram.
- In an automatic mode, fill the frame with the white paper and take a picture. Then fill it with the black sweater and take another. Look at both histograms.
- Set exposure compensation to +1 and photograph the white paper again, then +2. Set it to -1 and photograph the black sweater again, then -2.
- Put the paper on the sweater so both are in one frame, set compensation back to 0, and shoot once more.
How to judge it: in step 1, both graphs should sit near the middle and both pictures should look gray, which is the meter doing what it is built to do. In step 2, find the frame where the paper looks white with the graph close to the right wall but not climbing it, and the frame where the sweater looks black with texture still visible. Note those numbers; they are roughly how much to correct for snow and for dark scenes. In step 3, you should see two humps, one for each material, and you can check whether the paper has clipped. If you can predict where each graph will land before you look, you can read a histogram.
Questions about the histogram
What does a good histogram look like?
There is no single good shape. A good histogram is one that matches the scene and does not lose important detail at either edge. A sunny day often fills most of the width, while snow, night and silhouettes each give a different shape that is just as correct.
Should the histogram touch the right side?
Only if what touches it is something you can lose, such as a light source or a glint. Close to the right wall is good for noise; pressed against it across a large area is lost detail. For raw shooters, a little contact may still hold detail, because the camera graph is cautious.
Why does the histogram in my editor look different from the one on my camera?
The camera graph is made from its processed preview, and your editor measures its own rendering of the raw file. Different tone curves, color spaces and settings give slightly different graphs, so some difference is normal.
Why are there no numbers on the vertical axis?
Because each graph is stretched so its tallest peak fills the box. The heights show how common each tone is compared with the other tones in the same photo, not an absolute pixel count, so you cannot compare heights from one photo to another.
Does the histogram show focus or white balance?
No to focus: it counts brightness and nothing else, so zoom into the picture to check sharpness. The RGB view does show each color channel, but a strong color cast and a colorful scene can look alike on it, so judge white balance by eye or with a neutral reference.
What do gaps in the histogram mean?
A gap means no pixels have those brightness levels. A few gaps in a simple scene are normal. A regular comb of thin gaps after heavy editing means the tones were stretched, and it can show as banding in smooth areas such as skies.
Want to check what stuck without the game? The metering and histogram quiz asks about both, with an explanation for each answer.