A Bayesian Skin/Non-Skin Color Classifier using Non-Parametric Density Estimation
Faculty of Computing, Health and Science
School of Engineering and Mathematics / Centre for Communications Engineering Research
This paper addresses an image classification technique that uses a Bayesian decision rule for minimum cost to determine if a color pixel has skin or non-skin color. Our proposed approach employs non-parametric estimation of class-conditional probability density functions of skin and non-skin color with a feature vector that consists of all three components of the RGB color space. Experimental results demonstrate that the classifier can achieve good classification performance. Furthermore, its simplicity is an attractive feature for real-time applications. It is a useful tool for image processing tasks such as human face detection, facial expression and hand gesture analysis.