Abstract

In order to solve problems of uncovered regions in region-based motion estimation, we address a novel method which generates the uncovered region memory using backward motion estimation, and present results of the application to very low bit rate video coding. In the proposed algorithm, changed regions are detected by using frame difference (FD) and segment information. For the changed region of a previously reconstructed frame, backward motion compensated prediction is performed. Then, with motion information, the uncovered region memory is generated and updated without transmission of shape information. Whenever the uncovered region appears in the scene, the algorithm adaptively selects between spatial prediction from the previously reconstructed image and temporal prediction from a special uncovered memory. The regions with large prediction error are transformed and encoded. Simulation results show an improvement in the subjective and objective image quality due to the remarkable reduction of transmission bits for prediction error.

Full Text
Published version (Free)

Talk to us

Join us for a 30 min session where you can share your feedback and ask us any queries you have

Schedule a call