Motion vector estimation and encoding for motion compensated DWT

In this work, we propose a new technique for estimation and encoding of motion vectors, in order to achieve an efficient and scalable representation of motion information. The framework is Motion Compensated Three-Dimensional Wavelet Transform (MC3DWT) video coding. At low bit-rates an efficient estimation and encoding of motion information is especially critical, as the scarce coding resources have to be carefully shared between motion vectors and transform coefficients. The proposed technique, called Constrained Motion Estimation, outperforms the usual "unconstrained" one at low to medium rates, and is essentially equivalent to it at higher rates. Moreover, the proposed encoding technique for Motion Vectors, based on Wavelet Transform and context-based bit-plane coder, gives a scalable representation of them. © Springer-Verlag Berlin Heidelberg 2003