Abstract
In this thesis, we proposed a model for robust joint source and channel coding scheme for MPEG-2 video transmission. Instead of separate design of source and channel coding, we combine the source significance information with unequal error protection scheme in channel coding in order to minimize the overall distortion. Here we focus on MPEG-2 video transmission on noisy channel where the quantization of MPEG-2 compression introduces source distortion and the unavoidable channel errors bring channel distortion. The idea is to optimally distribute the source and channel bit rate so that the overall distortion will be minimized for a given total bit rate.The MPEG-2 encoded bit stream is split into three layers, I-pictures, P-pictures, and B-pictures, due to their error sensitivity and RCPC codes are used for efficient protection during channel transmission. We estimate the source and channel distortion function through simulation and show that joint source and channel coding with unequal error protection scheme can improve the system performance. To reduce the computation and employ the joint source and channel coding for real-time application, we found that the error propagation nature is tied up with the GOP structure of MPEG-2 encoded video and the models can be used for source and channel distortion estimation.