Extract Color Palette From a Photo

Extracting a palette from a photograph is a different job than extracting one from a logo or a flat design, because a photo carries continuous lighting and shadow rather than a small number of deliberate flat colors — the same red shirt can contribute a dozen subtly different swatches across its lit and shadowed sides before quantisation groups them back together. Six colors is enough headroom to capture that range — a couple of dominant tones, a supporting color, and one or two accents pulled from smaller but visually distinct areas — without the palette collapsing into six near-identical shades of the same hue.

A palette built this way from, say, a landscape or product photo tends to work well as the seed for a website theme or a design mockup: the most common swatch (usually the largest area, such as sky or backdrop) is a natural background or muted-surface color, and the smaller, higher-coverage-percentage swatches are natural candidates for accents specifically because the algorithm already favoured colors that are visually distinct from the rest of the image.

Worth checking the coverage percentage before committing a swatch to a specific role in a design: a color that covers 45% of the source photo will read as overwhelming if used as a small accent, while a color covering only 3% will disappear if used as a large background fill. Matching a swatch's coverage share to a role of similar visual weight in the new design usually produces a more balanced result than picking purely by which color looks nicest in isolation.