CIE Lab Colour Space and the Science of Skin Tone Measurement
The core technology of AI skin analysis is the CIE Lab colour space developed by the International Commission on Illumination (CIE). This device-independent, three-dimensional colour model measures all perceptible colours using three values: L* (lightness), a* (red-green axis), and b* (yellow-blue axis) (HunterLab, Konica Minolta).
CIE Lab is crucial in dermatology because it closely aligns with human visual perception and allows precise quantification of colour differences. The ITA (Individual Typology Angle) method, widely used for skin tone measurement, is calculated directly from CIE Lab. ITA computes the angle between L* and b* values to objectively classify skin colour from very light to dark tones (NIH/PubMed, University of São Paulo research).
ITA's significance lies in overcoming the limitations of the subjective Fitzpatrick Scale, which relies on self-reporting, by enabling objective, hardware-based measurements. While RGB colour space displays differently across devices, CIE Lab is device-independent, ensuring consistent analysis.
Browser-Based AI Analysis and Privacy
A notable approach in beauty tech is 'client-side AI,' where all analysis is completed within the user's browser. In this model, user-uploaded photos are never transmitted to external servers.
Technically, lightweight ML inference engines based on WebAssembly (Wasm) are utilised. Google's open-source MediaPipe framework can generate real-time facial meshes (468 landmark points) in browser environments, and TensorFlow.js runs trained models within the browser. This detects facial regions, converts the colour histograms of those regions to CIE Lab, and analyses skin tone.
This 'privacy-first design' is especially important for beauty and health services handling biometric data. As privacy regulations like GDPR (General Data Protection Regulation) strengthen, local computation methods that don't collect user data have become a key factor for building consumer trust. Services like haut.ai and L'Oréal Paris AI skin analysis are also evolving in this direction.