MacAdam Ellipses and Just-Noticeable Color Differences

A numerical difference is not automatically a visible difference

A just-noticeable difference is the smallest change that an observer can reliably detect under a defined condition. That threshold depends on the starting color, field size, surround, luminance, adaptation, viewing time, and observer.

Early chromaticity experiments showed that the region of indistinguishable colors around a test point is not a perfect circle and is not the same size everywhere. MacAdam ellipses visualize those local tolerance regions on the CIE 1931 xy diagram.

The ellipses expose nonuniformity

In some parts of xy space, a tiny coordinate change is visible. In others, a larger change is required. The ellipses also rotate, showing that sensitivity differs by direction as well as magnitude.

This is one reason raw xy distance is a poor universal color-difference metric. Perceptually oriented spaces and formulas were developed to make equal numerical distances correspond more closely to equal visual differences.

A one-step ellipse is not a universal production tolerance

MacAdam ellipses are sometimes described in steps, with larger multiples representing broader regions. A one-step claim does not guarantee invisibility in every application. Large uniform surfaces, repeated side-by-side products, gradients, and critical brand colors can reveal differences that pass a loose numerical threshold.

Conversely, texture, gloss, noise, small area, and complex surroundings can hide differences that an instrument reports clearly.

Delta E formulas improve comparison but still need context

CIELAB Delta E 76 uses Euclidean distance and is simple but uneven. CIE94 and CIEDE2000 add weighting for lightness, chroma, hue, and application behavior. Other spaces and formulas may be better for displays or HDR.

No single cutoff such as Delta E below two is universally correct. Acceptability depends on whether the task is textile lot control, display calibration, dental shade matching, paint repair, packaging, or interface design.

Build tolerances from the actual use case

Define the instrument, illuminant, observer, geometry, sample preparation, and formula. Separate average tolerance from maximum channel or component errors when needed. Include visual panels for high-value or effect materials.

A production tolerance is a business and perceptual decision informed by measurement. It should reflect what users can see and what the process can control.

What this means for digital design

In interfaces, small color differences should not carry essential meaning unless they remain clear across displays and viewing conditions. Use labels, shapes, spacing, and sufficient lightness separation.

HexCheck can calculate exact digital differences and contrast, but a passing metric should be treated as evidence, not the entire design decision.

A disciplined verification checklist

Record the difference formula, reference space, illuminant, observer, geometry, sample preparation, field size, surround, and acceptance purpose. Validate the threshold with representative samples rather than borrowing a number from another industry.

• Choose the formula that fits the application.
• Separate detectability from acceptability.
• Test side-by-side and isolated viewing.
• Include gloss, texture, and area in visual review.
• Use redundant cues when meaning is essential.

Continue in HexCheck