Skin Tone Analysis Through Skin Tone Map Generation With Optical Approach and Deep Learning

Area of Science:

• Dermatology and Image Analysis
• Optical Engineering
• Computational Imaging

Background:

• Traditional skin tone assessment methods like the individual typology angle (ITA) have limitations.
• ITA is sensitive to lighting conditions and does not accurately capture skin redness.

Purpose of the Study:

• To introduce an automated, image-based skin tone mapping method.
• To overcome the limitations of existing methods in assessing skin tone accurately.

Main Methods:

• Utilized optical approaches and deep learning for automated skin tone mapping.
• Developed a method to generate skin tone maps by leveraging illuminant spectrum and segmenting facial skin regions.
• Evaluated the method using simulated skin tone images under various standard illuminants (D45, D65, D85).

Main Results:

• Skin tone maps generated under D65 lighting conditions showed the highest accuracy (color difference ~6).
• The proposed method demonstrated a strong correlation between mapping positions and pigment levels.
• Effectively distinguished skin tones related to redness, outperforming the ITA method.

Conclusions:

• The automated method shows potential for improved skin tone assessment in cosmetic and medical fields.
• It mitigates the impact of illuminants and differentiates between dominant skin pigments.
• Further validation is needed, including measuring illuminant spectrum and physiological assessment.