Image Color Extractor

The image is scaled down to 200 px on its long side, similar pixels are grouped together, and k-means clustering in the Oklab color space, which follows how the eye perceives differences, finds the colors that best represent the picture. The same image and number of colors always give the same palette. A color that fills most of a group, such as the flat color of a logo, is reported exactly.

The palette favors colors that cover large areas, so a small detail, such as a red button in a big photo, can be absorbed by a neighboring color. Raise the number of colors, or click the detail with the eyedropper to get its exact code.

It is the share of the image closest to that color. Transparent pixels are left out, and the percentages add up to 100. A palette with fewer colors than you asked for means the image has fewer clearly different colors.

It reads the pixel under the pointer from the image at full resolution (images over about 16.7 megapixels are scaled down to that size first). In JPEG photos, neighboring pixels often differ slightly because of compression, so the magnifier helps you aim. Pick from screen, in browsers that support it, picks any color shown on your screen.

No. The image is read and analyzed in your browser with canvas and never leaves your device. Only the number of colors and the export format are remembered in this browser; the image and the picked colors are not.