A Probabilistic Model for the Human Skin Color

We present a multivariate statistical model to represent the human skin color. In our approach, there are no limitations regarding if the person is white or black, once the model is able to learn automatically the ethny of the people involved. We propose to model the skin color in the chromatic subspace, which is by default normalized with respect to illumination. First, skin samples from both white and black people are collected. These samples are then used to estimate a parametric statistical model, which consists of a mixture of gaussian probability density functions (pdf's). Estimation is performed by a learning process based on the expectation-maximization (EM) algorithm. In the following, experiments are carried out and receiving operating characteristics (ROC curves) are obtained to analyse the performance of the estimated model and compare it to outcomes of models that use a single gaussian density. Finally, conclusions are presented and future work is outlined. _________...

• Type: Conference
• Year: 2001
• Where: ICIAP
• Authors: Tibério S. Caetano, Dante Augusto Couto Barone
• Keywords: Gaussian Probability Density, ICIAP 2001, Image Processing, Multivariate Statistical Model, Parametric Statistical Model