A factor graph framework for joint source-channel decoding of images

We propose a framework for iterative joint source-channel decoding for communication with a fidelity criterion. We consider a class of source models that are used in current state-of-the-art transform image coding schemes. We construct a global graphical model that includes both the channel coding redundancy and the source model and we apply the sum-product algorithm to estimate the transmitted signal with minimum distortion. Our results show the promise of our framework for improving over existing techniques of digital communication.