Tracking human heads based on interaction between hypotheses with certainty

Abstract

We propose a method for tracking human heads, where interaction between hypotheses plays a key role. We model appearances of the human head and generate hypotheses for a human head in the image in the model space. We then propagate and reform hypotheses over time in turn to realize tracking human heads. During tracking, we bring about interaction between hypotheses to eliminate the hypotheses denoting false positives and, at the same time, to maintain the hypotheses denoting human heads. © Springer-Verlag 2003.