Three-dimensional lifting schemes for motion compensated video compression

Three dimensional wavelet decompositions are efficient tools for scalable video coding. In this paper, we show the interest of a lifting formulation of these decompositions. The temporal wavelet transform is inherently non-linear,d ue to the motion estimation step, and the lifting formalism allows us to provide several improvements to the scheme initially, proposed by Choi and Woods: a better processing of the uncovered areas is proposed and an overlapped motion compensated temporal filtering method is introduced in the multiresolution decomposition. As shown by simulations, the proposed method results in higher coding efficiency, while keeping the scalability functionalities.