数字图像处理第五章ME和MC技术要点解析.ppt

ME and MC Based on WT Yan Jingwen ME and MC Based on WT Video image motion estimation and motion compensation in DCT domain. Video image motion estimation and motion compensation in WT domain. ME and MC performance comparison in DCT and WT domain. ME of Multi-resolution level Motion Vector in different level Motion Vector Estimation Proceeded for Coarse to Fine Scale In Motion Vector in different level MultiResolution Motion Estimation (MRME) and Full Search Algorithm (FSA) are widely applied in video image compression. FSA have a high precision and a lot of time needed to obtain ME. Motion Vector in different level Because of the downsampling effect, the motion vector estimation in coarse scale gives rise to some error in comparison to the motion vector estimation in time domain, it make the predictive precision decrease. MPEG-1 Motion predicton MPEG-2 Motion predicton Interband self-similiarity Interband self-similiarity Motion prediction Using Interband self-similiarity 2D Motion Prediction(2DMP) Conclusion 2DMP can exploit the relations in different frames. Because of the limitation of searching area, 2DMP produces bigger errors when the scenes change speedily. 2DMP spend a volume of time, and a high compression ratio can be got. 3D SPIHT Compression 3D SPIHT can be derived from 2DSPIHT. 3D SPIHT can exploit 3D redundancy in a 3D video serial, so it does not need motion prediction. 3D SPIHT can take advantage of full data information in 3D data. 3D SPIHT Compression 3D video data construction 3D SPIHT Compression Alternatively apply wavelet transform along X,T, and T dimensions 3D SPIHT Compression 3D SPIHT Compression 3D SPIHT Compression 2D SPECK Compression Set Partition Embedded bloCK (SPECK) Coding Introduction Low Complexity Embeddedness Implementation easily 2D SPECK Compression 2D SPECK Compression 2D SPECK Compression 2D SPECK Compression 2D SPECK Compression 2D SPECK Compression 2D SPECK Compression 2D SPECK Compression 2D SPECK Compressio