Simple method of human skin detection using HSV and YCbCr color spaces
Muh. Arif Rahman, I Ketut Eddy Purnama, Mauridhi Hery Purnomo · 2014
Human skin detection is an important preliminary stage to improve the performance of other areas of object detection or recognition such as human face detection, hand gesture recognition, and pornography contents detection. Popular methods in this area are processing a single image pixel in HSV or YCbCr color spaces. The limitation of these approaches is they cannot address the wide range of the skin color distribution. This paper proposes a new approach by combining two model of skin color for each pixel into a vector contains color elements of H, S, Cb, and Cr. A set of experiments prove that the method produces an True Positif Rate (TPR) of 93.89% and False Positif Rate (FPR) of 10.75%. This result is significantly higher comparing those produced by single color models.