Image Color Palette Extractor - Get HEX, RGB & HSL

What the Color Palette Extractor Does

This tool answers a narrow question precisely: if you had to describe this image with a handful of colors, which ones would they be? It decodes your image in the browser, scales it down so the longest edge is about 200 pixels, and reads the raw pixel data off a canvas. Downscaling is not a shortcut that costs accuracy — averaging neighbouring pixels during the scale removes JPEG ringing and sensor noise that would otherwise show up as dozens of nearly identical near-blacks, and it keeps the whole analysis under about forty thousand pixels so it finishes instantly even on a phone. Pixels with an alpha value under 128 are discarded before anything else happens. That single step is what stops a logo on a transparent background from reporting black as its main color, which is the most common wrong answer other extractors give. The remaining opaque pixels are grouped with median cut. Median cut treats the pixels as a cloud in three-dimensional RGB space, wraps a box around them, repeatedly picks the box with the widest color spread weighted by how many pixels it holds, and slices it in two at a point where the color values actually change. Repeating that until there are as many boxes as you asked for gives clusters that follow the real structure of the image, and each cluster reports the average of the pixels inside it rather than an arbitrary bucket center. This is why the results are actual colors from your image instead of the washed-out greys that fixed-grid bucketing produces. The ordering is then adjusted. Ranking purely by pixel count is a trap: a photograph is often forty percent sky or white studio backdrop, and a raw count puts that first and buries the colors you actually wanted. Near-neutral, near-black and near-white clusters are therefore given less weight when the list is sorted, though they are never dropped and the percentage shown next to each swatch is always the true, unmodified pixel share. Finally, every swatch gets a text color computed from its WCAG 2.1 relative luminance, choosing black or white by whichever has the higher contrast ratio against that specific color. That test uses luminance, not lightness, so it gets mid-tones right where a simple brightness threshold flips to the wrong answer.