||This dissertation discusses several aspects and proposes improved methods of joint source-channel coding for efficient image and video data transmission over noisy channels. First, we propose a joint source-channel coding system for fractal image compression.
The system allocates the available transmission bitrate between the source and the channel coders using a Lagrange multiplier optimization technique and unequal error protection. Simulation results show that our method outperforms previous work in this field that only covered coding with a fixed-length fractal code.
Secondly, we discuss our findings with regard to the real-time aspect of newly emerging systems for the protection of embedded wavelet bitstreams against bit errors and packet erasures. Recently proposed algorithms for the distortion-rate optimization of channel coding rate assignments to different parts of the compressed bitstream are not suited to many real-time applications, since they require the operational distortion-rate function of the source coder with a rather time consuming computation. We propose the use of parametric models, instead of the true operational distortion-rate curves. We further propose a Weibull model of the distortion-rate curve, and show its superiority to the previous models for real-time applications. The Weibull model is used in two important joint source channel coding applications: Unequal error protection for the transmission of embedded image and video bitstreams over binary symmetric channels, and unequal loss protection for the transmission over packet erasure channels. Extensive simulations show, that using our parametric model instead of the true operational distortion-rate function, similar expected distortion is achieved, while, additionally, the real-time constraint is satisfied.
The third segment of this study discusses distortion-rate optimization of the progressive error protection of embedded codes. This is of utmost importance in I progressive transmission, where the reconstruction quality is important not only at the target transmission rate but also at the intermediate rates. Systems are considered that take successive blocks of embedded bitstreams and transform them into a sequence of channel codewords of fixed lengths using error detecting and correcting codes. We propose a real-time algorithm that searches for an error protection strategy that minimizes the average expected distortion over a set of
transmission rates. Experimental results for a binary symmetric channel show that our approach achieves more efficient results compared to currently known solutions when both reconstruction quality and time complexity are considered. If compared to the solution that optimizes the end-to-end performance of the system, the proposed scheme has a slightly worse performance at the target transmission rate and a better performance at most of the intermediate rates, especially at the lowest ones.
Finally, we propose a packet loss protection system for video streaming over the Internet. Our system is especially attractive for video bitstreams composed of a non-scalable base layer, and an embedded enhancement layer, such as the MPEG4-FGS bitstream. If embedded bitstreams are used, we assume that a minimum reconstruction quality should be guaranteed and, therefore, we treat a first part of the bitstream as the base layer, which should be completely decoded to get the required minimum video quality. The proposed system provides the strongest possible protection to the base layer using a hybrid ARQ scheme, and unequal packet loss protection to the enhancement layer. Experimental results show that our system yields a significantly improved performance over the widely used priority encoding transmission system.