December 2015
Spotlight Summary by Shakil Rehman
Human skin detection in the visible and near infrared
Distinguishing a human body from its surroundings through light reflection from skin is a formidable task in search-and-rescue operations and computer vision. The authors of this Applied Optics article propose a two-dimensional feature space to characterize human skin from the absorption properties of melanin, hemoglobin and water. In the case of skin, the scattering coefficient defines the reflectance at different colors, and absorption of various pigments gives the characteristic spectral features. The authors used a hyperspectral camera to record images and defined a 2D feature space that identifies skin from the background. The images are transformed into a reflectance-based skin index representing two infrared bands signifying the presence of melanin and water in the skin. Another index was defined for false alarm suppression, based on the green and red bands of the spectral reflectance on the premise that melanin and hemoglobin cause skin to be redder than most of the false alarm sources, which tend to be green. These two indices, when plotted against each other, define the two dimensional feature space, classifying skin from the background. The authors claim a superior performance of the proposed feature space in skin detection when compared to other color based features.
You must log in to add comments.
Add Comment
You must log in to add comments.
Article Information
Human skin detection in the visible and near infrared
Michael J. Mendenhall, Abel S. Nunez, and Richard K. Martin
Appl. Opt. 54(35) 10559-10570 (2015) View: Abstract | HTML | PDF