Fault-tolerant Error Correction For Video Encoding Transmission Optimization Method

Compressed video stream is highly sensitive to bit error. When it is transmitted in error-prone channel, picture is damaged too seriously to acceptable Wireless channel is a kind of error-prone channel. It is characterized by state variance and high error rate. How to increase error-resilience is one important work for wireless video communication application. In this thesis, we deep investigate the problem in video transmission, and propose four optimal error-resilient schemes.Traditional video rate control scheme has two disadvantages for error-prone network: first, the error correction coding makes rate control buffer overflow, and increases the number of skipped frames; second, the error correction coding uses too much bandwidth in high error rate channel. In order to resolve above two problems, one rate control scheme is proposed for wireless channel. The scheme is of these characters: joint error correction and video coding; hybrid FEC/ARQ error correction coding; optimal bit allocation between error correction coding and video coding by end to end rate-distortion formula.RPS (Reference picture selection) is one of effective schemes for video error-resilience. It selects appropriate reference picture to cease error propagation. RPS is of high complexity and low efficiency. In this thesis, an optimal low-complexity RPS scheme is proposed. Firstly, the encoder performs error concealment steps the same as the decoder on the basis of the feedback information. Secondly, the encoder selects an optimal reference picture between reconstructed error picture and preceding picture by proposed rate-distortion algorithm. In above process, picture error variance is predicted linearly to decrease the number of motion estimation process, and then computation complexity is lowered.MDVC (Multiple description video coding) is for multiple channels system. In this system, multiple equal important streams are transmitted in different channels. MDVC has one well-known disadvantage: Bit error causes the mismatch of reference picture and degrades the picture quality. To solve this problem, an error compensation MDVC scheme is proposed. In this scheme, error compensation information is transmitted also with generic description information. Description decoder decodes error compensation information to reconstruct the error picture. Additional, a bit allocation algorithm is proposed to adjust error compensation degree adaptive to the channel state.In erasure channel such as IP channel, error packet is whole discarded. In order to make full use of correct bits in error packet, multiple packets decoding scheme is proposed Firstly, through cross-layer optimization and bit level FEC coding, decoder could receive error packets. Secondly, by packet level FEC coding, decoder could decode multiple error packets.Experiments show all of the above four optimal schemes could improve picture quality greatly.