Source-channel coding and modulation for wireless image transmissions

Abstract

This thesis proposes new algorithms for image transmissions over wireless communication channels. The wireless image transmission is the first step in offering multimedia services over current wireless cellular networks. It is also a challenging problem because a wireless channel usually has a very high bit error rate and sometimes severe fading. The objective of this work is to achieve minimum end-to-end average distortion of the transmitted images. This work subsumes a number of results obtained by other researchers to date. In contrast to the other systems and algorithms, this work obtains a closed form bit allocation algorithm to minimize the average distortion for binary feedforward memoryless channels. This not only greatly reduces the computational expense of bit allocation, but also proves in theory that the optimal bit allocation is none increasing. The results are then extended to binary feedback channels, binary fading channels and power constrained channels. Two approaches from different directions achieve the efficient wireless image transmission. One approach applies joint source and channel coding to protect transmitted image information in given noisy channels. The other reduces the channel raw bit error rate directly by designing good spreading sequences in spreading spectrum systems. The first part of the thesis proposes bit allocation schemes for a variety of binary wireless channels. The so-called progressive transmission system is the main concern since it is very powerful and efficient for highly loaded communication channels. The second part of the thesis investigates the design and implementation of a new class of real-valued spreading sequences for Code Division Multiple Access (CDMA) systems. These new spreading sequences reduce the multiple access interference in multiuser spreading spectrum systems, and as a result, the average bit error rate of the channel as well. The performance gain of this new sequence class appears in both single user and multiuser detection schemes. In simulations, the proposed bit allocation algorithms give satisfactory image transmission under severe channel conditions and outperform previous joint source-channel coding systems in terms of the end-to-end PSNR measurement. The proposed new spreading sequences also manifest performance gain in various wireless channels, including multiuser and multipath fading channels.