MAWUPC algorithm for tracking face and hands in complex background

This paper proposes a new object tracking algorithm called the MAWUPC (motion adaptive weighted unmatched pixel count) algorithm for solving complex background and occlusion problems and for proposing a more effective combination of color and motion information. The proposed algorithm incorporates color transform, Kalman filter and UPC operation to make more effective a combination of color and motion information to track multiple objects of the same color. The proposed algorithm has advantages in rejecting static background objects of similar color without background training and subtraction, and in solving the problem of occlusion among target objects at the same time. This paper shows the efficiency of the proposed MAWUPC with an experiment of face and hand tracking in a complex background even when the hand occludes the face and when hands are crossed