Color Difference (Delta E) Calculator

About the Color Difference (Delta E) Calculator

Find out how different two colors really are. Enter a reference (the standard, target or proof) and a sample (the print, product or screen color you are checking) as HEX or any CSS color, as CIELAB L*a*b* values, or as LCh — and the calculator works out the four color-difference formulas in use today: ΔE00 (CIEDE2000), ΔE*94 with the graphic-arts and textiles constants, ΔE CMC(l:c) and the original ΔE*ab (CIE76).

It also tells you how the sample differs — lighter or darker, more or less saturated, redder, greener, yellower or bluer — places the result on a rule-of-thumb visibility scale, and checks whole lists of measurements against a tolerance in batch QC mode. The CIEDE2000 code is unit-tested against the 34 reference pairs published by Sharma, Wu and Dalal. Everything runs in your browser.

Common uses

• Checking a printed proof or press sheet against the target Lab values from a spectrophotometer.
• Setting and verifying color tolerances for packaging, textiles, plastics and paint.
• Comparing a brand color across files, apps and screens to see whether a change is visible.
• Choosing which of two near-identical design tokens to keep, or proving that two are effectively the same.

Which Delta E formula should I use?

ΔE00 (CIEDE2000) is the formula the CIE recommends for industrial color-difference evaluation (CIE 142, also ISO/CIE 11664-6). It corrects the way plain Lab distance overstates differences in saturated colors and handles blues and near-neutrals better. Use it unless your contract, standard or customer names another one.

• ΔE*ab (CIE76) is the straight-line distance in CIELAB. It is simple and still common in older specifications, but a ΔE*ab of 2 can be invisible in a saturated blue and obvious in a gray.
• ΔE*94 (CIE94) weights chroma and hue differences by the reference’s chroma. The graphic arts constants are kL = 1, K1 = 0.045, K2 = 0.015; the textiles constants are kL = 2, K1 = 0.048, K2 = 0.014.
• ΔE CMC(l:c) was developed for the textile industry and is standardised in ISO 105-J03. 2:1 is used to judge whether a match is acceptable; 1:1 to judge whether a difference is perceptible.

Numbers from different formulas are not interchangeable: a tolerance of 2.0 ΔE*ab and 2.0 ΔE00 are different requirements.

The formulas

All differences are sample − reference. C* = √(a*² + b*²) is chroma and ΔH* = 2√(C*₁C*₂)·sin(Δh/2) is the metric hue difference.

• ΔE*ab = √(ΔL*² + Δa*² + Δb*²)
• ΔE*94 = √((ΔL*/kL·SL)² + (ΔC*/SC)² + (ΔH*/SH)²) with SL = 1, SC = 1 + K1·C*₁ and SH = 1 + K2·C*₁ (C*₁ is the reference chroma)
• ΔE CMC = √((ΔL*/l·SL)² + (ΔC*/c·SC)² + (ΔH*/SH)²) with SL = 0.040975·L*₁/(1 + 0.01765·L*₁) (0.511 below L* 16), SC = 0.0638·C*₁/(1 + 0.0131·C*₁) + 0.638 and SH = SC·(F·T + 1 − F)
• CIEDE2000 first stretches a* for low-chroma colors (the G factor), then combines the lightness, chroma and hue differences with weights SL, SC and SH that depend on the average color, plus a rotation term RT that corrects the blue region: ΔE00 = √((ΔL′/kL·SL)² + (ΔC′/kC·SC)² + (ΔH′/kH·SH)² + RT·(ΔC′/kC·SC)·(ΔH′/kH·SH)). kL, kC and kH are 1 under reference conditions; textile work sometimes uses kL = 2 (set them under Advanced).

The implementation follows the notes and test data of Sharma, Wu and Dalal, Color Research & Application 30(1); CMC follows Clarke, McDonald and Rigg as given in ISO 105-J03.

D50 or D65?

CIELAB values are always relative to a white point. Graphic-arts measurement uses D50 with the 2° observer (ISO 13655), as do ICC profiles and CSS lab(); screen work and sRGB use D65. The same HEX color has different Lab numbers under the two white points, so compare data that share one.

HEX and other sRGB colors are converted with the sRGB definition (IEC 61966-2-1) and, for D50, the Bradford chromatic adaptation — the matrices and white points of CSS Color 4. Lab and LCh values you type are used as they are; the white point then only affects the swatches. CSS lab() and lch() are D50 by definition and are adapted when you work in D65; their hue may use deg, rad, grad or turn, and an alpha value below 1 is refused, as for other transparent colors.

Reading the direction of a difference

Positive ΔL* means the sample is lighter, negative darker. Positive ΔC* means more saturated (cleaner), negative duller. Positive Δa* means redder (less green), negative greener; positive Δb* yellower (less blue), negative bluer. ΔH* is the hue difference in the same units as ΔE*ab, positive when the sample’s hue angle is larger (counter-clockwise from the reference). The calculator writes these out in words, ignoring changes smaller than 0.1.

Batch QC format

Put one reference/sample pair on each line, with commas, semicolons or tabs between the values (copying cells from a spreadsheet gives tabs):

• two colors — #1d5fcb,#1f5fc8 or rgb(29 95 203);rgb(31 95 200);
• six numbers — L*, a*, b* of the reference, then of the sample (or L*, C*, h if you choose LCh);
• either of those after a name in the first column.

A first line with column names is skipped, and lines starting with # followed by a space are comments. With semicolons or tabs, numbers may use a decimal comma (50,5). Each pair passes when the chosen formula is at or below your tolerance; the report lists every pair with all four formulas and the CIELAB differences.

Limitations

• The visibility guide is a common rule of thumb for ΔE00 under good side-by-side viewing, not a standard. Real tolerances depend on the industry, the material, the size of the samples and the lighting — agree them with your customer or supplier.
• For HEX colors the tool compares the colors as sRGB defines them, not how a particular screen, printer or material reproduces them. To check physical samples, measure them with a spectrophotometer and enter the L*a*b* values.
• Lab values from different white points or observers (2° or 10°) must not be mixed in one comparison. HEX colors are converted with the 2° observer.
• CIE94 and CMC are not symmetric: they weight the difference by the reference color, so swapping reference and sample changes their result. ΔE*ab and ΔE00 do not change.
• CIEDE2000 can jump slightly for pairs on opposite sides of the color wheel, where the average hue is ambiguous — Sharma, Wu and Dalal describe these discontinuities (test pairs 9–12 show one).
• Batch QC reads up to 20,000 lines; the table shows the first 500 pairs and the downloaded report contains them all.

Privacy

Everything happens in your browser. What you enter or open here is not uploaded or stored by MySmartCoPilot.