Enhanced adaptive skin detection with contextual tracking feedback
Mehrdad Ghaziasgar, James Connan, Antoine Bagula · 2016
Skin detection in computer vision is the basis of many novel human-computer interface applications. The goal of skin detection is to accurately highlight skin pixels in an input image, while discarding all non-skin pixels. This makes it possible to accurately locate regions pertaining to a user in an input frame. A number of approaches to skin detection have been proposed over a number of years, with varying degrees of success. One computationally efficient method that provides the ability to dynamically adapt to users is based on histogram back-projection. This paper proposes a modified skin detection method based on histogram back-projection with key improvements that provide for improved skin detection results in a specific skin detection context.