Efficient representation of a video signal, the subject of video compression technology, is quite necessary due to its huge raw data size. Several standards are now available including H.261 and MPEG. This thesis investigates a more efficient video compression algorithm than the MPEG standard by combining 3-D subband/wavelet coding with motion compensation techniques.; Motion-compensated 3-D subband coding (MC-3DSBC) is proposed to overcome the limits of 3-D SBC and MC prediction based coding. It is novel both in its generation of 3-D subbands and in their subsequent encoding. First, 3-D subbands are generated by MC temporal analysis and a spatial wavelet transform, and then encoded by 3-D subband-finite state scalar quantizer (3DSB-FSSQ) exploiting inter-subband dependencies. The rate allocation from the GOP level to each class of subbands is optimized by utilizing the structural property of MC-3DSBC that the additive superposition rule approximately holds for both rate and distortion.; The optimization problem between the motion estimation and the quantizer is formulated and the best trade-off of rate between motion vectors and the compensated frame is investigated. A hierarchical variable size block matching (HVSBM) algorithm is presented to obtain a smooth as well as rate-constrained motion vector field. Optimal tree pruning is applied to an initial motion vector tree built by the hierarchical motion estimation with splitting. The operation of merging blocks, a key operation to VSBM, is performed in an optimal way.; The proposed algorithms were implemented in software and applied to several real-world test video sequences covering various camera and object motions. Experimental results illustrate their performances exceed those of previous methods including the MPEG algorithm when operated with equal bit rate while maintaining modest computational complexity and memory size.
展开▼
