Color difference

Delta E calculator CIEDE2000, CIE94, CMC and CIE76

Delta E puts a number on how different two colors look. Type two colors or two sets of Lab values, choose a formula, and see the difference, what it means and every step of the formula, checked against the published CIEDE2000 test data.

What you get

  1. Six formulas CIEDE2000, CIE94 for graphic arts and textiles, CMC 2:1 and 1:1, and CIE76, side by side.
  2. The formula, worked Every term of the formula you chose, computed for your two colors.
  3. Any input Hex, rgb(), hsl(), oklch() and names, or Lab values typed as three numbers.
  4. Checked All 34 test pairs Sharma, Wu and Dalal published for CIEDE2000, reproduced to four decimals.

The calculator

Measure two colors

Type two colors, or two sets of Lab values, and choose a formula. Every change is measured again at once.

The calculator needs JavaScript. The worked examples and the test pairs below read without it.

Color 1, the reference

#D3AF37

Lab 72.75, 0.95, 62.79

Color 2, the sample

#CDAA3D

Lab 70.92, 1.28, 58.42

The two colors, side by side CIELAB, D65 white

Result

CIEDE2000

1.84 ΔE00

Very close

Every formula, for these two colors
FormulaDifferenceReads as
CIEDE20001.84Very close
CIE94, graphic arts2.17Close match
CIE94, textiles1.44Very close
CMC 2:1, acceptability1.71Very close
CMC 1:1, perceptibility2.10Close match
CIE764.75Close match

The formula

ΔE₀₀ = √[ (ΔL′/S_L)² + (ΔC′/S_C)² + (ΔH′/S_H)² + R_T · (ΔC′/S_C) · (ΔH′/S_H) ]

Every term of the formula, worked
Every term of the formula, for these two colors
TermWhat it isValue
C*₁Chroma of color 162.7929
C*₂Chroma of color 258.4313
C̄*Mean chroma60.6121
GChroma correction for a*0.0005
a′₁Corrected a* of color 10.9477
a′₂Corrected a* of color 21.2823
C′₁Corrected chroma of color 162.7929
C′₂Corrected chroma of color 258.4313
h′₁Hue angle of color 1, degrees89.1353
h′₂Hue angle of color 2, degrees88.7425
ΔL′Lightness difference-1.8339
ΔC′Chroma difference-4.3616
Δh′Hue angle difference, degrees-0.3927
ΔH′Hue difference-0.4152
L̄′Mean lightness71.8347
C̄′Mean corrected chroma60.6121
h̄′Mean hue angle, degrees88.9389
THue weighting0.6108
ΔθRotation angle, degrees0.0000
R_CRotation chroma term1.9980
S_LLightness weight1.3209
S_CChroma weight3.7275
S_HHue weight1.5553
R_TRotation term for blues0.0000

The two colors

A CSS color, such as #D3AF37 or oklch(0.77 0.14 92), or Lab values as three numbers, such as 50, 2.68, -79.78.

Formula

Worked examples

Five pairs, four formulas

The same pairs, measured with four of the formulas. Open one in the calculator to see each formula worked.

Five pairs of screen colors: CIEDE2000, CIE94 for graphic arts, CMC 2:1 and CIE76, and how CIEDE2000 reads
PairColorsCIEDE2000CIE94CMC 2:1CIE76Reads asCalculator
Two light grays #E5E5E5#DCDCDC 2.00 3.18 1.11 3.18 Close match
Two saturated blues #0000FF#1E3CFF 7.30 7.93 7.59 22.67 Noticeable
Two yellows #FFD700#FFE135 3.29 3.53 3.21 8.36 Close match
Two near-blacks #121212#1E1E1E 3.58 5.80 5.67 5.80 Close match
A brand gold and a close match #D3AF37#CDAA3D 1.84 2.17 1.71 4.75 Very close
  • CIE76 overstates saturated colors. The two blues are 22.67 apart by CIE76 and 7.30 by CIEDE2000: the eye is less sensitive to chroma differences among strong colors, and only the newer formulas allow for it.
  • Neutrals are the other way round. The two light grays measure 2.00 by CIEDE2000 and 3.18 by CIE76; gray differences are easy to see, and CIEDE2000 stretches its lightness scale least near the middle.
  • CMC 1:1 is harsh on near-blacks. Its lightness weight is fixed below L* 16, so the two near-blacks measure 11.35 by CMC 1:1 but 3.58 by CIEDE2000.
  • A close brand match. The gold and its match are 1.84 by CIEDE2000, inside the tolerance most brand specifications set, while CIE76 puts them at 4.75.

Reference data

The 34 published test pairs

Sharma, Wu and Dalal published these pairs to catch the mistakes CIEDE2000 implementations make, most of them in how the hue angle and its mean are handled. This page reproduces every one to four decimals; the largest difference before rounding is 4.9e-5.

Sharma, Wu and Dalal (2005), table 1: each pair's Lab values, the published CIEDE2000 and this page's
PairLab, color 1Lab, color 2PublishedThis pageCalculator
150, 2.6772, -79.775150, 0, -82.74852.04252.0425
250, 3.1571, -77.280350, 0, -82.74852.86152.8615
350, 2.8361, -74.0250, 0, -82.74853.44123.4412
450, -1.3802, -84.281450, 0, -82.74851.00001.0000
550, -1.1848, -84.800650, 0, -82.74851.00001.0000
650, -0.9009, -85.521150, 0, -82.74851.00001.0000
750, 0, 050, -1, 22.36692.3669
850, -1, 250, 0, 02.36692.3669
950, 2.49, -0.00150, -2.49, 0.00097.17927.1792
1050, 2.49, -0.00150, -2.49, 0.0017.17927.1792
1150, 2.49, -0.00150, -2.49, 0.00117.21957.2195
1250, 2.49, -0.00150, -2.49, 0.00127.21957.2195
1350, -0.001, 2.4950, 0.0009, -2.494.80454.8045
1450, -0.001, 2.4950, 0.001, -2.494.80454.8045
1550, -0.001, 2.4950, 0.0011, -2.494.74614.7461
1650, 2.5, 050, 0, -2.54.30654.3065
1750, 2.5, 073, 25, -1827.149227.1492
1850, 2.5, 061, -5, 2922.897722.8977
1950, 2.5, 056, -27, -331.903031.9030
2050, 2.5, 058, 24, 1519.453519.4535
2150, 2.5, 050, 3.1736, 0.58541.00001.0000
2250, 2.5, 050, 3.2972, 01.00001.0000
2350, 2.5, 050, 1.8634, 0.57571.00001.0000
2450, 2.5, 050, 3.2592, 0.3351.00001.0000
2560.2574, -34.0099, 36.267760.4626, -34.1751, 39.43871.26441.2644
2663.0109, -31.0961, -5.866362.8187, -29.7946, -4.08641.26301.2630
2761.2901, 3.7196, -5.390161.4292, 2.248, -4.9621.87311.8731
2835.0831, -44.1164, 3.793335.0232, -40.0716, 1.59011.86451.8645
2922.7233, 20.0904, -46.69423.0331, 14.973, -42.56192.03732.0373
3036.4612, 47.858, 18.385236.2715, 50.5065, 21.22311.41461.4146
3190.8027, -2.0831, 1.44191.1528, -1.6435, 0.04471.44411.4441
3290.9257, -0.5406, -0.920888.6381, -0.8985, -0.72391.53811.5381
336.7747, -0.2908, -2.42475.8714, -0.0985, -2.22860.63770.6377
342.0776, 0.0795, -1.1350.9033, -0.0636, -0.55140.90820.9082

Pairs 7 to 16 test the hue angle and its mean around 0° and 180°; pairs 17 to 24 test large and small differences around one color; pairs 25 to 34 are real surface colors. Source: G. Sharma, W. Wu and E. N. Dalal, "The CIEDE2000 color-difference formula: implementation notes, supplementary test data, and mathematical observations", Color Research & Application 30(1), 2005, pages 21 to 30.

Reading the number

What a difference means

How this site reads a CIEDE2000 difference
CIEDE2000Reads asIn practice
Below 1ImperceptibleMost people cannot see it, even side by side.
1 to 2Very closeVisible to a trained eye, side by side.
2 to 5Close matchVisible side by side; the colors pass as the same apart.
5 to 10NoticeableVisible at a glance.
10 and upDistinctRead as different colors; at this size the formulas agree less.
  • A tolerance is an agreement. Brand and print specifications commonly allow 1 to 3 by CIEDE2000; textile work commonly uses CMC 2:1 with a tolerance near 1. The right number is the one the buyer and the maker agree on, for the formula they name.
  • Name the formula with the number. A difference of 2 means something else by CIE76 than by CIEDE2000, as the examples above show.

Method

How it measures

  1. A CSS color is read as sRGB, made linear, converted to CIE XYZ and then to CIELAB with the D65 white of sRGB screens. A see-through color is first blended over white.
  2. Three numbers are read as CIELAB values directly, as the published test pairs are given; the swatch shows the nearest sRGB color and says when the values lie outside it.
  3. CIEDE2000 follows CIE 142-2001 and ISO/CIE 11664-6:2014 with k_L, k_C and k_H all 1, implemented as Sharma, Wu and Dalal's notes describe.
  4. CIE94 uses k_L 1, K₁ 0.045 and K₂ 0.015 for graphic arts, and k_L 2, K₁ 0.048 and K₂ 0.014 for textiles; CMC uses l:c 2:1 or 1:1. Both weigh by color 1, the reference.
  5. The figures are also checked against two independent libraries, delta-e for CIE76, CIE94 and CIEDE2000 and Color.js for CMC, on 2,000 random pairs.

The worked examples and the test pairs are recomputed by an automated test from the same code, so a figure on this page cannot drift from what the calculator measures.

Found a wrong number? Tell us.

Limits

What a difference cannot tell you

  • The formulas model small differences. They were fitted to color pairs a few units apart; above about 10 they still rank pairs, but say less about how different the colors look.
  • The white point matters. This page uses D65, the white of sRGB screens; print work often uses D50, as CSS lab() does, and the same pair can then measure slightly differently.
  • Viewing conditions are assumed. The formulas assume two flat samples side by side under a set light; size, gloss, texture and the surroundings change what people see.
  • Two colors can match under one light and not another. That is metamerism, and a single difference under one white cannot show it.
  • Screens are not proofs. Colors typed here are screen colors; a printed or dyed sample has to be measured with an instrument.

Where next

More on measuring color

From one difference to the science behind it.