Fast and accurate object tracking by succesive learning of lifting wavelet filters

This paper presents a novel object tracking method which uses lifting wavelet filters. Our method is based on successive learning of free parameters in a lifting filter. To realize fast learning, the free parameters are determined by solving some minimization problems using Lagrange method. The filter learned for the object in the previous frame is used as an initial filter of a new lifting filter to be learned for the object in the present frame. In simulation, human right eye in a sequence of images is tracked accurately by our method.