Abstract

In this thesis, the error resilient transmission of H.264 compressed video using Context-based Adaptive Binary Arithmetic Code (CABAC) as the entropy code is examined. The H.264 compressed video is convolutionally encoded and transmitted over an Additive White Gaussian Noise (AWGN) channel. Two iterative joint source-channel decoding schemes are proposed, in which slice candidates that failed semantic verification are exploited. The first proposed scheme uses soft values of bits produced by a soft-input soft-output channel decoder to generate a list of slice candidates for each slice in the compressed video sequence. These slice candidates are semantically verified to choose the best one. A new semantic checking method is proposed, which uses information from slice candidates that failed semantic verification to virtually check the current slice candidate. The second proposed scheme is built on the first one. This scheme also uses slice candidates that failed semantic verification but it uses them to modify soft values of bits at the source decoder before they are fed back into the channel decoder for the next iteration. Simulation results show that both schemes offer improvements in terms of subjective quality and in terms of objective quality using PSNR and BER as measures. Keywords: Video transmission, H.264, semantics, slice candidate, joint source-channel decoding, error resiliency