Description
Judge whether two colours are perceptibly different by converting them to a perceptually uniform space such as CIELAB or u'v' and comparing their distance with a just-noticeable threshold.
Colour matching experiments show trichromacy and near-linear matching, which underpin linear spaces such as CIE XYZ. But just-noticeable differences, found by altering a colour until observers can only just tell it changed, form MacAdam ellipses whose size varies strongly across the CIE xy diagram, so raw xy distance is a poor guide. A uniform colour space is one where coordinate distance fairly reflects perceived difference, so distances below a threshold mean colours cannot be told apart. CIE u'v' improves chromaticity uniformity, and CIELAB, which includes lightness, is the most widely used. The check is calculable from colour values alone. Forsyth and Ponce warn that comparing small differences is useful but comparing large differences is usually dangerous, and perceived colour also depends on surround, illumination and adaptation.
Attributes
- Discipline
- Cognitive Psychology & Perception
- Loss type
- Siloed
- Stage
- Design, Evaluation
- Data
- Analytical / expert
- Effort
- Hours
Steps
- Express the colours to compare in device-independent coordinates (convert sRGB to CIE XYZ with the display white point).
- Convert them to a uniform space: CIE u'v' for chromaticity only, or CIELAB to include lightness.
- Compute the Euclidean distance (Delta E) between each pair.
- Compare the distance with a just-noticeable threshold (a Delta E of about 2-3 is commonly treated as just noticeable) to judge whether users could tell them apart.
- Use the check only for small differences; do not rank large differences against each other.
- Confirm critical distinctions under realistic surrounds and displays, and test with colour-vision deficiency simulations.
Sources
- Computer Vision: A Modern Approach (Forsyth & Ponce, 2011)
Also known as
- CIE LAB distance
- Delta E check
- Uniform colour space distance
- MacAdam ellipse threshold
Categories
- User Research
- Cognitive Psychology & Perception