Joint source-channel (de)coding for progressive image transmission using turbo and LDPC codes

Progressive image transmission has recently undergone enormous development and is now being used currently in applications such as web browsing and image transmission between mobile devices. The corresponding embedded compressed bitstreams are subject to severe error sensitivity and propagation problems when being transmitted over noisy channels, especially over wireless links. In addition, to apply the forward error correction (FEC) techniques, joint source-channel coding and/or decoding methods are employed to increase the overall system performance.; In this thesis; a joint source-channel coding (JSCC) system is proposed for transmitting JPEC2000 encoded images. The procedure has separate design methods for the image header and data packets. Two hybrid FEC/ARQ schemes operating over an AWGN channel to realize near error-free transmission of the image header are studied. Compressed data packets for different quality layers are protected using an unequal error protection (UEP) method. A set of rate compatible punctured turbo (RCPT) codes is used for FEC. A theoretical rate-distortion based channel rate allocation model, as well as a practical channel segmentation method are established. The practical method is based on channel operating regions and avoids complicated computation needed for optimization. A joint source-channel decoding (JSCD) system is also proposed using the JPEG2000 image compression standard and LDPC codes for both AWGN and flat fading channels. A regular LDPC code is used as the FEC. The log-domain iterative sum-product algorithm is chosen as the channel decoding method. A general modification algorithm was designed to update the log-likelihood ratio (LLR) of partial codeword bits after each iteration. A framework for a density evolution analysis is given to show that it is a theoretically valid procedure to track the error messages for the joint decoding system.; Simulation results show that the JSCC UEP method reduces the computational complexity of the rate assignment procedure and also outperforms the EEP scheme by up to 6 dB when the source coding rate is 1.0 bpp. The JSCD method reduces the channel decoding time in the range 2.43% to 16.25%. The improvement in quality of the reconstructed images can be up to 3 dB compared to systems not employing a joint decoding design.