Error concealment for low bit rate video communication

Abstract

Error concealment (EC) is an efficient way to improve video quality against transmission errors because compressed video bit streams are extremely sensitive to packet loss over error-prone channels. Once packet loss occurs, the decoder may not reconstruct the video image well, and even fail to decode the bit streams if important data are missing. We study the problem of recovering pixels of lost macroblocks by using the existence and directions of neighboring pixels. In this paper, first, we propose a solution by detecting the directions of neighboring macroblocks and then predict the lost pixels by using directional weighted interpolation. Second, combing edge detection and directional interpolation algorithm, we propose an adaptive spatial-temporal algorithm scheme to deal with problems caused by abrupt scene changes of video sequences. The experimental results show that this method can improve the subjective quality of corrupted frames and obtain higher PSNR compared with algorithms in the H.264/AVC Joint Model decoder.