A NOVEL TEMPORAL ERROR CONCEALMENT ALGORITHM BASED ON RESIDUAL ESTIMATION

摘要

The temporal error concealment is an important post processing technique to combat transmission errors for video communication, which is usually achieved by motion vector estimation and temporal replacement using temporal correlation existing in video data. However, the quality of concealed video is not satisfactional due to missing residual data lost in transmission. In this paper, a novel temporal error concealment algorithm based on residual estimation is proposed. The motion compensated temporal concealment is used to recover the most damaged macroblock firstly. Then the residual is estimated by the boundary residual of current macroblock and predictive macroblock. Lastly the estimated residual is added to the restored macroblock to achieve more accurate video quality. Experimental results show that it significantly outperforms other existing temporal error concealment algorithm and can efficiently improve the video quality by 2 to 5 dB PSNR in average.