Image Histogram Viewer

How the Image Histogram Viewer works

Image histogram viewer: what to know first

A histogram is the most honest picture of an exposure. Screens lie: a photo looks brighter on a phone at full brightness than on a laptop in daylight, and a dark room makes everything look well exposed. The histogram does not change with the screen. It counts every pixel by its brightness, from 0 for pure black on the left to 255 for pure white on the right, and the height at each point is how many pixels have that value. That is why every serious camera, raw editor and video monitor shows one.

How to read an image histogram

The shape tells you where the tones of the picture sit, not whether the picture is good. A night scene should crowd to the left and a snowfield to the right. What matters is whether the shape matches the scene, and whether anything is jammed against an edge.

Clipped detail cannot be recovered by editing a JPEG. Brightening a crushed shadow just turns black into flat grey. Photographers call the habit of keeping the curve as far right as possible without clipping "exposing to the right", because it captures the most tonal detail with the least noise, then pull the brightness down afterwards.

The three views

RGB overlaid draws the red, green and blue channels on top of each other in additive colour. Where all three coincide the curve looks white; where they separate you can see a colour cast. A blue curve sitting to the right of the others means a cool image, and red to the right means a warm one, which is the usual result of tungsten light.

Luminance converts the image to grey first, weighting green most and blue least in the same way the eye does, and shows a single curve. It is the one to read for overall exposure.

Separate channels stacks luminance and the three colours as four plots, so none of them hides another. Use it to catch a single clipped channel.

Vertical scale switches between a linear count and a logarithmic one. The log scale squashes the tallest peaks and lifts the sparse tails, so a small amount of clipping or a thin spread of highlight detail becomes far easier to see.

Reading the stats

• Source is the size of the image analysed. Large images are scaled to 1600 pixels on the long side before counting, which does not change the shape of the distribution.
• Mean RGB is the average of each channel, with 128 as the midpoint. If one channel's mean is well above the others, the image has a cast in that colour. For example, 142 / 128 / 104 is a warm image, typical of indoor light or a golden-hour sky.
• Std dev RGB measures spread, which is a practical proxy for contrast. Values in the single digits or low teens mean a flat, foggy image; 60 or more is punchy, high-contrast material.
• Luminance gives the same two figures for the grey version, the best single summary of brightness and contrast.
• Near black / white is the share of pixels within eight levels of either end. A note appears when either passes 2 %, the usual sign of crushed shadows or blown highlights.

Where it helps

• Product photos on white backgrounds should show a tall spike at 255 for the background and no spike just below it, which would mean a grey, not white, backdrop.
• Scans and screenshots that look washed out usually show a gap at both ends; stretching the levels to fill it restores contrast.
• Comparing edits: run the before and after, or put them side by side with image compare, to confirm an edit added contrast rather than just clipping.

Pitfalls

• A single channel can clip while luminance looks fine. Red flowers, sunsets and stage lighting often blow out the red channel alone, so check Separate channels before trusting the near-white figure, which is measured on luminance.
• Graphics are not photos. A logo or illustration shows a few sharp spikes at its flat colours. That is correct, not a problem.
• Compression moves the curve. Heavy JPEG compression smears fine tonal steps. Check the histogram before and after following how to compress an image for the web.

The histogram can be downloaded as a PNG. Everything runs in your browser, so the photo is never uploaded.

Related tools