IMPROVING TEMPORAL ERROR CONCEALMENT BY GRNN IN VIDEO COMMUNICATION
Jun Horng Chen, Shih Chun Shao, Wen Hui ChenAbstract
This work aims to improve the temporal error concealment for the corrupted macroblocks whose motions are not locally-smooth. It is demonstrated that the recovered quality by the oft-used motion estimation approaches is not visually satisfied for those MBs of which adjacent MBs do not have a consistent movement. Therefore, this work will propose and demonstrate that, if the conventional error concealment approach is followed by the nonparametric regression approach GRNN, the concealed quality will be raised. The simulation results will show the proposed approach indeed improves the performance of error concealment and the improving gain is about 1 dB of PSNR.
Read Submission [902]