Methodology

The same formulas run in the ColorID app and on this site, so a color reads the same in both.

Reading a color

The loupe averages a 5 × 5 pixel area. The sRGB value is converted to CIELAB (D65 white point) with the standard sRGB transfer function.

Matching

Distances are CIEDE2000 (ΔE2000). For the nearest color in a catalog the 30 closest colors by plain LAB distance are checked with ΔE2000. Alternatives in other brands are one color per brand, at most six, within ΔE 6.

How close is close

Under ΔE 2: very close, hard to tell apart on a wall. 2 to 4: close, visible side by side. 4 to 6: similar, a substitute.

LRV

Published LRV is shown where a brand publishes it. Otherwise LRV is CIE relative luminance Y × 100, marked with ≈.

Color blindness simulation

Simulation uses the Machado, Oliveira and Fernandes (2009) matrices at full severity on linear RGB. Correction (daltonize) adds back the lost difference along the axis that is still seen. Achromatopsia keeps luminance only.

Under different light

The warm bulb and neutral light previews scale linear RGB by the white point of a 2700 K and a 4000 K light relative to 6500 K daylight. Eyes adapt to light, so the real shift is smaller; the preview shows the direction of the change.

Palettes

Photos are scaled to 150 pixels on the long edge; up to 10,000 pixels are clustered with k-means. Clusters closer than 30 RGB units are merged, colors under 1% are dropped.

Plain names

Plain names are a set of 31 names chosen by lightness, chroma and hue ranges in CIE LCh, built for color blind readers.

Limits

A photo carries the camera's white balance and the light in the room, and a screen adds its own shift. Every result is a guide; a physical sample is the reference.