Delta E Color Difference (CIEDE2000)

What it is for

"Is this close enough?" comes up whenever a color is reproduced: a logo red on a new printer, a paint match, a product photo against the brand swatch, two design tokens that might be duplicates. Delta E (ΔE) turns that question into a number. A difference around 1 is at the edge of what most people notice side by side, many workflows accept 2–3, and above 5 colors look clearly different.

This tool compares one reference with up to 50 samples. You can type HEX codes, CSS names, rgb(), oklch() or measured lab() values. Each row shows four metrics: CIEDE2000 (the current CIE recommendation), CIE94 with graphic-arts weights, CIE76 (plain Lab distance) and the OKLab distance used by CSS gamut mapping. You choose which metric and tolerance decide pass or fail, and you can download the table as CSV.

It does not measure colors itself; values must come from your files or an instrument.

Worked example

A brand red is #C8102E. Four candidate prints were scanned as #C51230, #B8162F, #D0122C and rgb(200 16 46). With CIEDE2000 and a tolerance of 2 the tool reports: Within tolerance: 2 of 4 samples.

The third sample shows why the metric matters: by old CIE76 it is 4.38 away, which sounds like a clear miss, while CIEDE2000 puts it at 2.14 — just over the line. CIE76 overstates differences in saturated colors; CIEDE2000 corrects for that. The last sample is the same color written in another notation, so every metric is exactly 0.

For a lab-measured check, the second example uses pair 1 of Sharma's published CIEDE2000 test data: lab(50 2.6772 -79.7751) against lab(50 0 -82.7485) gives 2.0425, the published value.

How to read the result

• The summary and metrics count passes and failures and show the largest ΔE2000.
• The swatch strip puts each sample under the reference. Out-of-gamut values are drawn with their CSS-mapped HEX, but the numbers use the original color.
• OKLab × 100 is the OKLab distance multiplied by 100, so 2 equals the one-JND step CSS uses.
• The CSV has every metric and the pass flag for your QC records.

Tips and common mistakes

• Agree on the metric before the tolerance: "ΔE 2" means something different in CIE76 than in CIEDE2000.
• CIE94 is not symmetric: the first color is the reference. Keep the reference fixed when comparing batches.
• Lab values here use the D50 white, like CSS lab() and most print workflows. Convert D65 measurements first.
• HEX values are already rounded to 8 bits; one step in a dark channel can be worth ΔE2000 0.4.
• For on-screen tokens, ΔE2000 below 2 usually means two tokens can be merged.

FAQ

Which Delta E should I use?

CIEDE2000 for most work; it is the current CIE recommendation and corrects CIE76's errors in saturated and blue colors. Use CIE76 or CIE94 only to match an older spec.

Is the CIEDE2000 implementation verified?

Yes. It reproduces all 34 test pairs published by Sharma, Wu and Dalal (2005) to four decimal places.

Why is the difference between my two HEX colors not zero although they look identical?

Below about 1 ΔE2000 most viewers cannot see a difference, but the numbers still differ.

Can I compare colors with transparency?

The alpha channel is ignored; only the opaque color is compared.