Date of Award
Doctor of Philosophy (PhD)
Li, Dr. Robert Y.
Video is one of the most powerful forms of multimedia because of the extensive information it delivers. Video sequences are highly correlated both temporally and spatially, a fact which makes the compression of video possible. Modern video systems employ motion estimation and motion compensation (ME/MC) to de-correlate a video sequence temporally. ME/MC forms a prediction of the current frame using the frames which have been already encoded. Consequently, one needs to transmit the corresponding residual image instead of the original frame, as well as a set of motion vectors which describe the scene motion as observed at the encoder. The redundant wavelet transform (RDWT) provides several advantages over the conventional wavelet transform (DWT). The RDWT overcomes the shift invariant problem in DWT. Moreover, RDWT retains all the phase information of wavelet coefficients and provides multiple prediction possibilities for ME/MC in wavelet domain. The general idea of variable size block motion compensation (VSBMC) technique is to partition a frame in such a way that regions with uniform translational motions are divided into larger blocks while those containing complicated motions into smaller blocks, leading to an adaptive distribution of motion vectors (MV) across the frame. The research proposed new adaptive partitioning schemes and decision criteria in RDWT that utilize more effectively the motion content of a frame in terms of various block sizes. The research also proposed a selective subpixel accuracy algorithm for the motion vector using a multiband approach. The selective subpixel accuracy reduces the computations produced by the conventional subpixel algorithm while maintaining the same accuracy. In addition, the method of overlapped block motion compensation (OBMC) is used to reduce blocking artifacts. Finally, the research extends the applications of the proposed VSBMC to the 3D video sequences. The experimental results obtained here have shown that VSBMC in the RDWT domain can be a powerful tool for video compression.
Suliman, Ahmed Abdelgadir, "Variable Block Size Motion Compensation In The Redundant Wavelet Domain" (2011). Dissertations. 135.