Research On Error Concealment Algorithms Based On H.264 Decoder

With the rapid development of multimedia and network, the video have won the favor of people due to it's lively and abundant content. Usually, the huge video data will be compressed before transmission, and the compressed streams are very sensitive to the channel error. One bit error may deteriorate the quality of both the current frame and the successive frames in the same GOP. Therefore, the error concealment technology will be a important topic with video communication.Compared with the former video coding standard, H.264/AVC has higher compression efficiency, more friendly network and stronger error concealment. In this paper, the basic principles of H.264 are described firstly. Then kinds of error control technologies are introduced. In the end, we analyze primary error concealment algorithms detailedly in H.264 decoder, such as spatial error concealment, temporal error concealment and frequency error concealment.For the shortage of original spatial and temporal error concealment, we apply the official H.264 test model JM86 to improve them. First of all, this paper proposes an adaptive spatial error concealment algorithm based on edge direction entropy. In the algorithm, we can calculate the edge direction information entropy to select bilinear interpolation method or directional interpolation method adaptively. In order to improve the edge quality, the traditional directional interpolation method has been improved, and a new method with combination of gradient amplitude and the number of edge points is presented. Finally, the paper also improved the origiral temporal error concealment algorithm by dividing macroblock into four sub-blocks whose motion vectors are predicted respectively, and then carry through error concealment to the imaged picture by means of the introduction of new boundary matching algorithm. Experimental results show that, compared with the original algorithm, the improved temporal and spatial error concealment algorithm provide better objective and subjective quality.