Return to search

Joint source-channel coding for image transmission and related topics

With the integration of wireless technologies and multimedia services, transmitting high quality images and video has become one of the main objectives for next generations mobile network systems. Shannon's classic separation theorem states that, under ideal conditions, source coding and channel coding can be treated separately without sacrificing any performance for the whole system. However, this theorem holds true only under ideal conditions. Practical communication systems do not meet such requirements. Therefore, joint source and channel coding may reduce distortion, as well as complexity and delay. In this thesis, different schemes of joint source-channel coding and decoding for error resilient image transmission over noisy channels are examined. Unequal error protection (UEP) is one of the techniques used in joint source and channel coding. A JPEG image is partitioned into DC components and AC components according to their respective sensitivity to channel noise. The highly sensitive DC components are better protected with a low coding rate, while the less sensitive AC components use a high coding rate. Simulation results show that the proposed UEP scheme slightly outperforms conventional equal error protection (EEP). A novel turbo diversity scheme (TDS) applied to JPEG coded images is proposed. Turbo codes have a built-in structure that is suitable for diversity techniques used to improve the quality of communications over a multi-path channel. The same image data is encoded by two separate turbo encoders and sent over two independent channels. The received data is then passed to a single turbo decoder. By utilising the built-in structure of the turbo encoder, the transmitted JPEG data encoded by a rate half code is recovered at the receiver using a more powerful rate third code yielded by the TDS. An iterative source-channel decoding scheme applied to JPEG coded images is investigated. Huffman codes used as the variable-length coding scheme in JPEG coding can be represented by an irregular VLC-trellis structure. State transition probabilities can be derived from the irregular trellis and can be used as a priori information to help iterative decoding between source and channel a posteriori probability (APP) decoders. Iterative decoding of JPEG coded images only gives a small coding gain due to the poor distance property of the original JPEG Huffman codes. We propose to replace the Huffman codes used in JPEG coding with error resilient source codes with larger free distance. After accounting for the penalty due to the increased average codeword length, the new scheme achieves a 4 dB coding gain over the conventional system for a range of SNRs. While the focus of this thesis is on joint source-channel coding, two other related topics are also examined, namely, capacity and normalisation of intersymbol interference (ISI) channels and parallel data convolutional codes. Previously published results showed a minimum Eb/N0 of -4.6 dB, 3 dB below the capacity of a flat channel, is obtained using the water-pouring capacity formulas for the 1+D channel. However, these results did not take into account that the channel power gain can be greater than unity when water-pouring is used. We present a new generic power normalization method of ISI channel frequency spectra, namely peak bandwidth normalisation, to facilitate the fair capacity comparison of various ISI channels. A final contribution presented in this thesis regards the proposed parallel data convolutional codes (PDCCs). The encoder inputs consist of the original block of data and its interleaved version. We propose a novel single self-iterative soft-input/soft-output (SISO) decoder structure for the decoding of PDCC. It has the advantage of needing only one APP decoder. Although the performance is not very encouraging, the novelty of the self-iterative idea behind the design is worth exploiting. / thesis (PhDTelecommunications)--University of South Australia, 2003.

Identiferoai:union.ndltd.org:ADTP/173379
Date January 2003
CreatorsXiang, Wei
Source SetsAustraliasian Digital Theses Program
LanguageEnglish
Detected LanguageEnglish
Rights© 2003 Wei Xiang

Page generated in 0.0021 seconds