Efficient adaptive-shape partitioning of video
Multimedia Tools and Applications
Block-Based Compressed Sensing of Images and Video
Foundations and Trends in Signal Processing
Hi-index | 0.00 |
This paper points out a major weakness of the conventional motion-compensated video compression paradigman assumption that the encoded information is sufficient to describe correct motion. We propose a new paradigm, based on multi-hypothesis expectation (MHE). Not only can conventional methods be reinterpreted as special cases of the MHE approach, but the new model suggests improved methods with provably superior performance. Blocking artifacts can be eliminated and prediction error reduced with no change in the encoder's motion search and no extra side information. The use of state variable conditioning can further enhance performance. Additional benefits are obtained when increased encoder complexity is allowable. Search algorithms and coding methods developed previously for block matching can be used with the new techniques with no change in bit-stream syntax. Mean-square prediction improvements in the range of 18% to 40% (0.9 to 2.2 dB) are obtained using MHE methods of varying complexity.