Hands detection and tracking for interactive multimedia applications
Résumé
The context of this work is a European project art.live which aims at mixing real and virtual worlds for multimedia applications. This paper focuses on an algorithm for the detection and tracking of face and both hands of segmented persons standing in front of a camera. The first step consists in the detection of skin pixels based on skin colour: the HSI and YCbCr colour spaces are compared. The colour space that allows both fast detection and accurate results is selected. The second step is the identification of face and both hands among all detected skin patches. This involves spatial criteria related to human morphology and temporal tracking. The third step consists in parameter adaptation of the skin detection algorithm. Several results show the efficiency of the method. It has been integrated and validated in a global real time interactive multimedia system.
Origine : Fichiers produits par l'(les) auteur(s)
Loading...