Y. Suzuki, C.S. Boon, and T. Horikoshi (Japan)
Video coding, Resolution enhancement, Signal processing, Inter prediction, Block segmentation.
This paper proposes a new approach for conducting high precision prediction of pixels in a target block based on a corresponding reconstructed block at lower resolution. Many previous approaches perform fine prediction between frames by extending the range and precision of motion parameters, such as block size and sub-pel motion vectors, resulting in the increase of coding bits for motion parameters at areas with detailed texture and/or motion features. In our approach, 1) the data of a decimated block which is the decimated version of the target block is first predictively encoded, then 2) the pixels in the target block are predicted by performing fine prediction between the reconstructed decimated block and previously reconstructed frame. We incorporated the proposed prediction scheme into a video coding algorithm using conventional motion compensated prediction to investigate the advantage of our approach. Simulation results show that the proposed prediction method improves the coding efficiency of the conventional coding scheme up to 4.7%.
Important Links:
Go Back
