Skip to Main Content
In this paper, we considers designing source-optimized irregular repeat accumulate (IRA) codes. The idea is to take advantage of the capability of unequal error protection (UEP) of IRA codes against errors because of their irregularities. Most existing UEP codes only provide different error rates for information bits of different importance but fail to achieve best performance for a given criterion (e.g. averaged distortion). However, in this paper, we propose an approach to design IRA codes optimized for such a criterion. We assume a cost function for the outcomes of the decoder and the error events, among the outcomes, which are determined by the location of the first error bit. We incorporate this cost function into channel code design and obtain IRA codes that are optimized over a new cost function instead of probabilities of error. We design source-optimized IRA codes with inherent UEP capabilities for the SPIHT-coded bitstream as an example of this general idea and use them for scalable image transmission. Conventional ways to achieve UEP are to design rate compatible punctured channel codes and UEP schemes separately. Using the distortion-rate function of the source coder as the cost function and the density evolution in an IRA code design, the optimization of channel codes and the choice of UEP schemes are embedded in the design process of IRA codes. Our design thus achieves source-optimized channel coding or true joint source-channel coding. In simulations, it shows that our scheme outperforms all previous UEP schemes and our results are only 0.40∼0.73 dB away from the theoretical limit.