A fast MRME scheme for video compression based on coefficients masking in the wavelet domain

A high efficiency multiresolution motion compensation (MRME) based on the wavelet representation is presented. The wavelet transform decomposes a video frame into a set of subimages with different frequency bands. In this approach, motion vectors in higher resolution are predicated by the motion vectors in the lower resolution and are refined at each step. In order to further increase the efficiency of motion compensation, a masking module which indicates the significant regions will be inserted between the modules of wavelet decomposition and the MRME. The significant regions are determined by an adaptive strategy based on the relationship between the wavelet coefficients of the current video frame and reference frames. Experiments shows that this new approach has a superior performance in terms of the processing time and the signal-to-noise ratio.