Tracking of objects in image sequences using multi-freeness mean shift algorithm

Current mean shift tracking algorithm uses a single radius parameter to describe the scale of the target. Each target has the position and size freeness only. This is not suitable for complex movements of the objects. In this paper we provide a new algorithm in which a bandwidth matrix is employed to describe the objects with two directions determined independently. Target-angle is also introduced to accommodate the rotation of objects. Furthermore we bring forward an efficient search strategy to cope with a temporal occlusion of the objects. Experimental results show that the new algorithm is able to adapt to any kind of object's movements and therefore has better tracking precision.