Highly scalable video coding by bidirectional predict-update 3-band schemes

Motion-compensated temporal filtering is an essential ingredient of recently developed wavelet-based scalable video coding schemes. The lifting implementation of these decompositions represents a versatile tool for spatio-temporal optimizations and numerous improvements have thus been proposed. An additional feature has been introduced lately by providing a 3-band lifting temporal decomposition, allowing non-dyadic scalability factors. The paper presents a motion-compensated scheme which is the 3-band equivalent of the 5/3 temporal filterbank. The proposed structure does not enter the classical lifting framework and therefore its invertibility is studied. Simulation results show higher coding efficiency, improved visual quality and more flexibility compared to the previously introduced Haar-like 2 or 3-band schemes.