Robust image and video transmission across heterogeneous networks with packet erasures and bit errors

In this dissertation, a rate allocation algorithm is developed, using a trellis structure and the Viterbi search algorithm. The rate allocation algorithm couples a source coder and a channel coder, thus providing joint source and channel coding, by allocating unequal error protection (UEP) to the channel packets. The rate allocation is sub-optimal in an expected mean squared error (MSE) sense, but the results indicate the sub-optimality is sufficiently small so as to be ignored in most cases. The complexity of the rate allocation system is significantly less than any known globally optimal search. The allocation process requires knowledge of the distortion-rate function of the source coder, and the packet error probabilities of the channel coder. The distortion-rate function, or an approximation thereof, for the source coder can often be obtained directly from the source encoder with minimal complexity increase. In the work presented, the error in the distortion estimates from two source encoders (JPEG-2000 and 3-D SPIHT) are shown to be greater contributors to sub-optimality than the rate allocation algorithm.; Using the rate allocation process, this dissertation presents a system for robust transmission of images and video across noisy channels. The binary symmetric channel model is used for the image transmission work as well as for general Gaussian and Laplacian sources. The channel model is extended to include erasures for the video transmission work. Simulation results demonstrate the system performance, having a high expected peak signal to noise ratio (PSNR), and is robust to both random errors and erasures.