Reading an Image Histogram, Spotting Clipping, and Comparing Versions
A screen lies to you. Its brightness, its calibration and the light in the room all change how a photo looks, which is why every serious editing application puts a histogram on screen. The Histogram Analyzer gives you the same measurement in a browser tab, with nothing uploaded, and two companion tools help you verify what your edits actually did.
What a histogram is and is not
A histogram is a tally, not a picture. The horizontal axis runs from pure black on the left to pure white on the right, and the height of the graph at each point is the number of pixels in the image with that value. It tells you nothing about where those pixels are. A photo of a black cat on snow and a photo of a chessboard can produce similar graphs.
There is also no correct shape. The internet is full of advice suggesting a nice hill in the middle is the goal, but a photograph of fog is supposed to bunch in the midtones, a night scene is supposed to pile up on the left, and a high-key portrait on a white background is supposed to lean right. The histogram tells you what you have; only you know what you intended.
Luminance and the three channels
The tool draws two charts. The RGB chart plots red, green and blue separately. The luminance chart plots perceived brightness, calculated with the standard Rec. 709 weighting - green contributes about 71 percent, red about 21 percent, blue only about 7 percent - because human vision is far more sensitive to green light than to blue.
Both charts matter, and they fail in different ways. An image can look perfectly exposed on the luminance chart while a single channel is already jammed against the right edge. That is colour clipping: a saturated red flower where the red channel has maxed out loses all its petal texture even though the overall brightness is fine. Always glance at the RGB chart before deciding an image is safe.
How to spot clipping
Clipping means pixels have been pushed past the ends of the scale and merged into a single value. On the graph it looks like a vertical spike welded to the very edge of the chart, as opposed to a curve that simply slopes down towards the edge and ends there.
- Blown highlights pile up at the right edge. Those areas are pure white with no recorded detail. No slider recovers them, because there is nothing to recover.
- Crushed shadows pile up at the left edge, turning shadow detail into flat black.
- Not all clipping is a mistake. Specular reflections off chrome or water, the sun itself and a genuinely black night sky are supposed to clip. What matters is whether the clipped region is a subject or a highlight.
Combing means you have stretched too far. A histogram with regular gaps between spikes, like the teeth of a comb, is the signature of an eight-bit image whose tones have been pulled apart by heavy contrast or exposure edits. The gaps are values no pixel has any more, and they show up in smooth areas such as skies as visible banding. When you see combing, go back and make a gentler edit rather than fixing it downstream.
Other shapes worth recognising
| What you see | What it usually means |
|---|---|
| Everything bunched in the middle, edges empty | Low contrast - haze, a dirty lens, or a flat scan |
| Two distinct peaks | A bright background and a darker subject, or vice versa |
| Data stops well short of the right edge | Underexposure, or a black point that never reaches white |
| One channel offset from the other two | A colour cast - a warm indoor light or a cool shade |
Comparing before and after
Numbers tell you what changed; your eyes decide whether it improved. The Image Diff tool loads two images onto one canvas and wipes between them with a slider that follows your pointer, showing the first image to the left of the line and the second to the right. Two habits make it far more revealing: load the two versions at identical dimensions so nothing shifts as you wipe, and move the slider back and forth rather than parking it. Motion is what makes small differences visible - a slight sharpening halo or a compression artefact that is invisible in a static side-by-side jumps out when it appears and disappears repeatedly under a moving edge. It is the fastest way to judge whether a smaller export is still good enough, or whether an upscale actually added detail or just softness.
Averaging many frames into one
The third tool takes a different approach to analysis: instead of comparing frames it merges them. The Image Average tool sums the pixel values of every image you load and divides by the count, accumulating in floating point so the intermediate sums stay precise.
That single operation does several useful things. Random sensor noise is different in every frame, so it averages towards zero while the real scene stays put, which turns a burst of noisy handheld night shots into one much cleaner image. Moving elements blur into smooth streaks, which is how you fake a long exposure of water or traffic without a neutral density filter. And people walking through a scene are in a different place in every frame, so a stack of shots from a fixed viewpoint can leave the location looking nearly empty. All of it depends on the frames being aligned, so shoot from a tripod or a steady surface and keep the framing identical.
Practical notes
These tools read whatever your browser can decode - JPEG, PNG, WebP and similar - so convert a camera RAW file to an eight-bit format before analysing it, and remember that the histogram then describes the converted version, not the raw sensor data. Analysis is also a good moment to check the rest of the file: the article on what EXIF data reveals covers the capture settings and location tags riding along beside the pixels. And if you find the idea of reducing an image to brightness values interesting, ASCII art is the same measurement turned into an output medium.
Because everything is computed on a canvas in your own browser, you can analyse client work, medical imagery or unpublished photography without any of it leaving your machine.