Get the color palette from an image

The obvious way to find an image's colors is to count how often each exact value appears and take the most common. On anything photographic that fails badly.

A blue sky is not one blue. It is thousands of almost identical blues, each appearing a handful of times, so each one scores close to nothing. Meanwhile a flat background that really is a single value wins easily, and you get a palette of the least interesting part of the picture.

This groups similar colors together before counting, so those thousands of near-identical blues become the one blue a person would point at and name. That is why the result matches what you see rather than what the file technically contains.

Each swatch shows how much of the image it accounts for, and the biggest share is very often a background rather than a brand color.

A logo on white will report white as dominant, which is true and useless. The useful signal is usually second and third. Treat the percentages as a sanity check on what you are looking at, not as a ranking of importance.

If you extract from a JPG or a screenshot, expect to be very close rather than exact, and it is worth knowing why.

JPG compression changes pixel values, most of all near edges and in flat areas. A screenshot carries whatever color profile the display was using. And anti-aliasing blends the artwork into its background, creating in-between shades that were never part of the design at all.

Extract from a flat PNG exported from the source and the match is normally exact. If the values really matter, they live in the vector file or the brand guide, and no tool reading a picture can beat the original.

Most brands work with three to five. One dominant color, one or two supporting it, and a neutral for text and backgrounds.

This tool does not ask you how many to find, which is deliberate. Nobody knows that number before they look, and tools that ask will happily split one color into four shades to hit whatever you typed. What comes back here is however many the image actually has, decided by how far apart the colors are rather than by a number you guessed.

So a photograph returning nine and a logo returning three is not the tool being inconsistent. It is the honest answer in both cases.

The hex and RGB values here are precise for screens. What this cannot give you is CMYK or Pantone, and you should be cautious of free tools that claim to.

Converting a screen color to print is a decision involving paper stock, ink and press behavior, not a formula. It is why printers ask you for the values rather than deriving them. Use these as the starting point and have the conversion done properly for anything that will be printed.

A brand guide normally records all four, which is hex for the web, RGB for digital, CMYK for print and Pantone for anything that has to match exactly across suppliers.