Hyperspectral Imaging Database of Human Facial Skin

Area of Science:

• Biomedical Optics
• Computer Vision
• Human Perception

Background:

• Human skin color perception is influenced by lighting and skin's spectral reflectance.
• Understanding skin spectral properties is crucial for accurate color reproduction and analysis.
• Existing datasets may lack the resolution or scope for detailed facial skin analysis.

Purpose of the Study:

• To develop and validate a comprehensive database of hyperspectral images of human faces.
• To provide a resource for applications in psychophysics, object recognition, and material modeling.
• To characterize spectral reflectance variations across different skin tones, sexes, and facial locations.

Main Methods:

• Acquired hyperspectral imaging data (400-720 nm, 10 nm steps) from 29 human faces.
• Collected data under controlled lighting and facial movements.
• Analyzed spectral reflectance within and between nine facial regions, validated with point/contact spectral measurements.

Main Results:

• Demonstrated that spectral reflectance profiles vary significantly between different skin tones, individuals, and facial positions.
• Identified substantial local variations in spectral reflectance across the face.
• Highlighted the limitations of average values from conventional measurement devices.

Conclusions:

• The validated hyperspectral face database is a valuable resource for diverse scientific applications.
• Facial skin spectral reflectance is highly heterogeneous, necessitating detailed, pixel-level analysis.
• Caution is advised when using averaged spectral data due to significant local variations.