Image Histogram

Builds a 256-bin histogram for the red, green, blue, and luminance channels (using Rec. 709 weights), then plots them at linear or log scale. Useful for spotting clipping at black or white, evaluating exposure, comparing colour casts between two photos, or checking that a generated image hits the dynamic range you expected. Switch to a single channel to see per-channel statistics (mean, median, min, max).

The two ends of the plot are where the useful information sits. A spike hard against bin 0 or bin 255 means pixels were clipped, and clipped pixels have lost their data rather than merely being dark or bright: no amount of later adjustment recovers detail from a region that recorded a single value. A histogram that stops well short of either end, by contrast, means the range is unused and the image can be stretched without losing anything.

Log scale exists because the linear one hides small counts. A few hundred blown highlights in a multi-megapixel image are invisible next to a peak of hundreds of thousands, so a picture that looks clean on the linear plot can still have clipped areas. Switch to log when you are looking for the presence of something rather than its quantity.

Luminance here is the Rec. 709 weighted sum, which is the standard for sRGB and HD video and weights green far above blue because the eye does. It is not the same as the average of the three channels, and it is not perceptual lightness either, so a histogram centred at 128 does not mean the image looks mid-grey.

The counts describe the decoded pixels the browser produced, which is not necessarily what the file holds. A raw file rendered through a different converter, an image with an embedded profile the browser applied, or a photo downscaled for display will all produce different histograms from the same original. Feed it the actual export you care about rather than a thumbnail.