A New Method for Skin Color Classification Based on Global CIELAB Data and k-Mean Clustering

Abstract

Human skin color varies significantly across the globe, presenting a challenge for accurate and objective classification. Existing systems, like the Fitzpatrick Skin Type (FST) and individual typology angle (ITA), are primarily designed to assess photosensitivity and fail to capture the full spectrum of skin color variations. This paper presents a novel method for skin color classification based on spectral measurements and k-means clustering. Data from 2770 women across four continents was collected using a spectrophotometer, measuring skin reflectance to calculate CIELAB color space values. Applying k-means clustering, six distinct skin color clusters were identified, offering a more objective and comprehensive classification system. This method provides a valuable tool for various fields, including cosmetics, dermatology, and computer vision, enabling the development of more inclusive products and applications.