Application of shift-invariant wavelet transform to video coding
Mohammed A. Al‐Mohimeed, Ching-Chung Li · Proceedings of SPIE, the International Society for Optical Engineering/Proceedings of SPIE · 1997
The standard discrete wavelet transform lacks translation invariance in 1-D signals and 2-D images. The down-sampling at each coarser scale accentuates the undesirable effects of the shift-variance, in particular, on the motion estimation from decomposed subimages in video coding. In this paper, we present a study of applying the Chui-Shi shift-invariant wavelet transform using 'oversampling frames' to video compression. Further, we present an algorithm for approximating the motion fields at different scales and different frequency bands by utilizing the multiresolution structure of wavelet decomposition. Motion vectors at a higher resolution are predicted by the motion vectors at a lower resolution through a proper scaling. Experimental results on a salesman video sequence show that the use of the 2-D oversampling algorithm of a biorthogonal spline wavelet has reduced the required number of motion vectors while maintaining an acceptable prediction error when compared to the classical block matching technique using the standard wavelet transform. The proposed approach will advance the video compression methodology for applications to HDTV and video conferencing.© (1997) COPYRIGHT SPIE--The International Society for Optical Engineering. Downloading of the abstract is permitted for personal use only.