Skin color detection using PCA-based color representation

Skin color detection is the process that classifies unknown colors into skin or non-skin classes. Skin color detection is preliminary step for facial or gesture analysis and can reduce search space for next higher level processing. In this paper, we show that proposed PCA-based color representation can give better performance than other frequently used color spaces such as XYZ and Luv. For skin detection, we use two classification models - histogram model and elliptical boundary model from skin and non-skin colors. The experimental results show the PCA-based color representation is more efficient than other color representation and the existing PCA-based representation based on the two classification models. © 2014 SERSC.