Boundary Artifact Minimization on Best Matching Blocks in Wavelet-Based Video Compression

The widespread usage of digital videos calls for an intense need of video compression techniques to conserve storage space and minimize bandwidth utilization. Discrete wavelet transform (DWT) has manifested its intrinsic multiresolution and scalability advantages on image compressions without the annoying blocking artifacts. Thus, Motion estimation and motion compensation (MEMC) combined with wavelet transformation has been beneficially applied to improve video compression efficiency and accuracy. However, circular convolution in wavelet transform generates border distortion because of the discontinuous periodic extension on the two ends of the signal, and becomes more apparent as more levels of decomposition are involved. In addition, hierarchical motion vector refinement needs accurate motion estimation from the lower resolution subframes without further distortions after multiple convolutions with the wavelet filter. This paper investigates the use of symmetric-extended wavelet transform (SWT) to reduce the boundary distortions experienced in multiresolutional motion estimation, so as to improve the compression performances in scalable video applications.