Research Of Error Concealment Techniques In Wireless Video Transmission Based On H.264

With the rapid development of video compression technology and network communication technology, especially the wireless video communication, the transmission and processing of video information-centered multimedia data has blossomed out as the focus both in wireless communication and multimedia field. However, due to random errors and burst errors brought by the multi-path effect and fading in wireless channels, the video quality decreases greatly. In addition, the existing video compression standard has pursued the improvement of video compression bitrate for target, so that video streams are more sensitive to the errors. Once there is transmission errors occur, the spread of errors will lead the reconstruct image to a sharp decline. Therefore, error concealment technology in wireless video transmission is significant to ensure the video transmission quality, and has become a hotspot in video communicantion field.Based on the analysis of H.264 video standard and phenomenon such as errors, packet loss and error diffusion in video transmission over wireless channels, this dissertation has conducted detailed research focusing on error concealment technology in wireless video transmission. The poor performance of existing spatial error concealment algorithms in complex texture areas is pointed, as well as the loss of a whole frame during transmission over wireless channels. Through the study of the structural characteristics of the image areas, and motion vector estimation of missing frame, the major innovations finished are as follow:1. Against the problems of image blurring in complex texture areas using existing spatial error concealment algorithms, an effective spatial error concealment method based on tensor voting is proposed, which extracted structural information by using tensor voting. Experimental results show that the proposed method improves the recovery accuracy effectively, especially in the region with more complex texture, and comparing with traditional algorithms, PSNR of concealed image using proposed method is improved more than 2dB.2. On the basis of improving the recovery accuracy of lost macroblocks with complex texture, an improved spatial error concealment algorithm which combined with the traditional algorithms is further proposed. The lost marcoblocks have been classified into different categories, and then an appropriate concealment method has been chosen to recover the lost macroblocks. Experimental results show that the subjective and objective quality of concealed image is improved greatly by using proposed algorithms.3. Against the condition of missing whole frames which occur frequently in low bit-rate packet-based wireless video transmission, an improved algorithm based on hybrid motion vector extrapolation is proposed, which uses not only the forward extrapolated MVs, but also the backward extrapolated MVs. In this way, more accurate MVs of the pixels in the missing frames are obtained. Then according to the motion characteristics of video sequence, a strategy which could reduce the computational complexity of the algorithm is brought, and a suitable error concealment method for whole frame loss is chosen by the motion intensity of video. Experimental results show that this proposed algorithm is highly effective and significantly outperforms other existing frame recovery methods with low computation.