Research On Video Error Concealment Based On H.264

The technology of the video compression, making use of intra and inter frame prediction coding techniques and variable length entropy coding techniques, can remove a large amount of redundant information. However, the transmission channel instability will usually lead to error and packet loss in the process of video transmission and makes the compressed code stream’s ability to fight errors very fragile. Once error takes place, it will be spread on the temporal and space, caused a sharp decline in the quality of video. In order to obtain a better video quality, it is meaningful to have error concealment in the decoding.In this thesis, we carried out the further research on video error concealment technique of H.264/AVC. Using of de-noising technique and spatio-temporal adaptive error concealment techniques, we improved the quality of video decoding. This article mainly studies the following aspects:1) We analyzed the reasons of high computational complexity of error concealment based on Non-Local Means denoising(NLM). First, the search window can be narrowed down by using the characteristic that the similarity of any two image blocks diminishes with their Euclidean distance. Afterwards, using norm inequality,we proposed a acceleration algorithm to reject the low similarity candidate pixels’s calculation of weight coefficient by judging whether the L1 norm of the block is less than the given threshold associated with temporal error concealment quality.2) For the shortage of the error concealment algorithm of H.264/AVC reference model. We proposed a match criterion to combine SSIM with BMA. Furthermore, we proposed a spatio-temporal adaptive error concealment algorithm based on the macroblock division. The experiment shows that proposed algorithm can further improve the values of PSNR and SSIM, and the quality of the video of error concealment can match to subjective feeling of human vision well.