Lterative Joint Source-Channel Decoding Using Turbo Codes For H.264 Video Transmission

Compression is always used on video signals when they are transmitted due to bandwidth considerations. Compressed video is fragile in that a few channel errors may result in severe quality degradation because of error propagation, synchronization loss, etc. This phenomenon is especially correct for H.264 video stream, for its highest compress ratio among all video compressing schemes. To alleviate this, channel coding is usually applied to protect the compressed video. Since their introduction by Claude Berrou, etc in 1993, turbo codes have been shown to give very good error correction performance. It could provide near Shannon limit performance in terms of bit error rate. In addition, compressed video still contains residual redundancy which can be used to combat channel noise. In this paper, an iterative joint source-channel decoding using turbo codes for MPEG-4 video transmission is investigated.This paper presents a novel iterative joint source-channel decoding scheme for H.264 video transmission over noisy channels. The proposed scheme, on one hand, utilizes the channel soft outputs generated by the turbo decoder to assist video decompression. On the other hand, the syntactic/semantic information from the video decompressor is used to modify the extrinsic information so as to improve the error correction capability of the turbo decoder. Simulation results show significant improvement in terms of PSNR, reconstructed video quality, as well as BER over turbo decoding only.This paper is organized as follows:Chapter 1 gives the background and meaning of joint source-channel decoding schemes.Chapter 2 gives the overview of several video compressing standards, and chief techniques used therein.Chapter 3 gives detail description of H.264 video compressing standard, including techniques of dequant, inverse transform, intra-prediction, inter-prediction and deblocking. Simulation results show its incomparable compressing performance.Chapter 4 introduces Turbo code, including its encoder and decoder, with special attention on MAP decoder algorithm. Simulation result shows that its performance in terms of Bit Error Rate (BER) is close to the Shannon limit.In Chapter 5,first, for data corruption, I propose to combine error control coding together with error concealment. Classical error detection is used to locate errors in the compressed video stream. This information, together with information provided by the syntax of the compressed bitstream is used to perform the error concealment. And a new concealment method which relies in the redundancies provided by the syntax of the compressed bitstream is introduced. Then, for cell loss in packet switched network, I proceed to generate candidates for corrupted slices and test their syntactical integrity and their special smoothness. The smoothest candidate for a slice which is syntactically correct is chosen as the correct slice. The improvement in PSNR performance is shown. At last, I proposed the iterative joint source-channel decoding scheme for H.264 video transmission over noisy channels. The scheme combines the syntax information of compressed video stream and the extrinsic information of Turbo decoder, which could improve the anti-error performance of H.264 decompressor while decreasing the iteration number of Turbo decoder. And this chapter also gives the simulation results of the proposed scheme, in terms of PSNR, reconstructed video quality, as well as BER.