How palette extraction works
Read more
A photo can contain millions of distinct pixel values, most of them near-duplicates, so the extractor simplifies before it counts. The image is scaled down to a small grid, each pixel's channels are quantized so near-identical shades fall into the same bucket, and the buckets are ranked by population. Each bucket's swatch is the average of the real pixels inside it, which keeps the result true to the photo instead of snapping to artificial grid colors.
Ranking alone would fill the palette with five nearly identical shades of the dominant color, so the extractor also enforces distance: a candidate only makes the cut if it is visually distinct from the swatches already chosen. The result reads like a designer's summary of the image, with the dominant tones first and the strong accents surviving even when they cover little area. Click any swatch to load it as the current sample, then copy it or carry it into the full picker.