Research On Key Techniques Of Video Error Concealment Methods Based On H.264

With the fast development in Internet technology, video technology has more and more applications such as digital TV, Internet streaming, mobile TV, video conference and so on. Due to the limitation of bandwidth and improvement on resolution, video compression and reliability in transmission is still the main problem in video transmission. Recently the H.264 standard can achieve quite high compression ratio, but as the compression rate goes higher, the video bit-stream is more fragile in face of channel errors, so the error control techniques is very important.Error concealment method is an error recovery technique in the decoder side, which does not require extra bandwidth. This paper mainly focuses on temporal error concealment method and frame loss error concealment method.FMO dispersed coding is an anti-error technique in the encoder side, if efficient error concealment algorithm is added in the decoder side, the anti-error performance will be further enhanced. A hybrid three-step error concealment method is proposed in this thesis based on traditional methods. It splits macroblocks with four neighboring macroblocks correctly received into small blocks and uses interpolation and RWBMA together to conceal different kinds of small blocks. Test results show that the proposal outperforms traditional algorithms in dealing with the bit-stream of FMO dispersed coding.In severe cases of channel errors or low bit video transmission scenarios, frame loss could happen. A frame loss bidirectional error concealment method based on MV field coherence is given in this thesis based on traditional methods. First, the implementation details of changing decoding flow to achieve bidirectional concealment are demonstrated while it is roughly mentioned before. Then a selection method based on MV field coherence is designed based on traditional methods to better choose the reference MVs after extrapolation. Simulation tests prove that this method is better than some traditional methods in terms of both objective quality and subjective quality.