Skin detection using statistics of small amount of training data
M. Shoyaib, Mohammad Abdullah-Al-Wadud, Oksam Chae, R. Byungyong · Electronics Letters · 2012
A skin detection approach based on combination of the statistics of multiple sources is presented. As long as the scarcity of available training data (with ground truth) is very common considering practical applications, such a fusion offers a much better classification of skin pixels compared to the state-of-the-art methods. Experiments on a renowned dataset result in a similar decision.