Multiple description source coding using forward error correction codes

Abstract

We present an efficient multiple description (MD) source coding scheme to achieve robust communication over unreliable channels. In contrast to the popular signal processing based methods, we propose channel coding principles to correlate the descriptions, and then use this correlation for combating channel impairments. We propose a fast, nearly optimal algorithm that aims to maximize the expected quality at the receiver given the channel state and the side channel rates. Our scheme can be used in conjunction with any source coder that is scalable, and is most easily matched to coders outputting a progressive bitstream. It has applications to the transmission of audio, images, as well as delay-constrained video signals, and can also be used to achieve reliable multicast transmission over the existing Internet with the use of simple protocols. Comparisons of our scheme on standard test images to some of the existing state-of-the-art signal processing based MD methods suggest that our simple scheme outperforms them by significant margins.