Read an image histogram

This is a histogram viewer for images: it counts where every pixel sits across the 256 tonal levels and draws the distribution as luminance, as red, green and blue overlaid, or as three stacked channel panels. Underneath the curve it reports the mean level, the median level, and the share of pixels pinned at pure black and at pure white — the two readings that say whether a photograph is merely dark or actually unrecoverable. It writes nothing back to the image; there is no export except the readings as text.

How it works

  1. Drop an image Drop or browse for one image. The pixels are counted in the tab, and the file name, size, format and dimensions appear in the File panel.
  2. Choose a view Luminance draws one Rec. 709 weighted curve and answers whether the image is exposed. RGB draws the three channels over each other and Split stacks them in three panels; both answer which channel clipped.
  3. Switch the scale if the shadows are flat Linear scales to the tallest interior bin. Logarithmic compresses the counts so a few hundred crushed pixels lift off the baseline instead of staying invisible.
  4. Read the four numbers Mean and median are given out of 255. Clipped shadows and clipped highlights are percentages, and turn amber past 0.5% and red past 5%, with a matching bar drawn at that edge of the chart.
  5. Copy the readings Copy the readings puts the pixel count, mean, median, both clipping figures and the one-sentence exposure verdict on the clipboard as plain text.

Questions

Can I edit the image from the histogram page?
No. The histogram is a reading tool: it counts pixels and draws the distribution, and the only thing you can take away is the readings as text. To act on what it tells you, use a tool that writes an image back, such as levels or white balance.
Does it measure every pixel of a large photo?
Up to four million pixels, yes. Above that the image is sampled down first, using nearest neighbour rather than smoothing, so every value counted is a real pixel from the original and the clipping fractions stay exact. A notice states the sample size and the original dimensions when this happens.
What counts as a clipped pixel?
A pixel whose luminance lands on level 0 or level 255. Those are the two ends of the eight-bit range, where detail is gone and no edit brings it back. Fully transparent pixels are skipped entirely and left out of the total, so an unused corner cannot put a phantom spike in the curve.

Every tool