Motion estimation with the redundant wavelet transform

We present a fast method for computation of the motion flow based on the redundant Haar transform of reference frames in a video sequence. This algorithm gives a multiscale representation of the motion flow, which allows progressive encoding. The high efficiency of the algorithm is achieved by a multiscale correlation between the current and reference frames of the coefficients on each of the Haar blocks. The algorithm is designed to adapt dynamically to make optimal use of channel capacity.