Tracking Skin-Colored Objects in Real-Time
A Argyros Antonis., Manolis I.A. · Pro Literatur Verlag, Germany eBooks · 2005
In this paper, a method for tracking multiple skin-colored objects has been presented. The proposed method can cope successfully with multiple objects moving in complex patterns as they dynamically enter and exit the field of view of a camera. Since the tracker is not based on explicit background modeling and subtraction, it may operate even on image sequences acquired by a moving camera. Moreover, the color modeling and detection modules facilitate robust performance in the case of varying illumination conditions. Owing to the fact that the proposed approach treats the problem of tracking under very loose assumptions and in a computationally efficient manner, it can serve as a building block of larger vision systems employed in diverse application areas. Further research efforts have focused on (1) combining the proposed method with binocular stereo processing in order to derive 3D information regarding the tracked objects, (2) providing means for discriminating various types of skin-colored areas (e.g. hands, faces, etc), (3) developing methods that build upon the proposed tracker in order to be able to track interesting parts of skin-colored areas (e.g. eyes for faces, fingertips for hands, etc) and (4) employing the proposed tracker for supporting human gesture interpretation in the context of applications such as effective human computer interaction. Acknowledgements This work was partially supported by EU IST-2001-32184 project ActIPret.