Improved quantization in multiple description coding by correlating transforms

The objective with multiple description coding (MDC) is to code one source of data into multiple bitstreams. The coding is done in such a way that multiple levels of quality is achieved. This means that even if one or a few of the bitstreams are lost, the received bits should make it possible to get an approximated version of the original data. One way to do this is to use pairwise correlating transforms which introduce correlation between the bitstreams. This correlation can be used in order to get an estimate of a lost stream. In this paper a new approach for MDC using pairwise correlating transforms is presented. In this approach, contrary to previous work, quantization of the source data is performed after the data has been transformed. This makes it possible to improve the shape of the quantization cells and to tailor these to the employed transform. We demonstrate that this offers a substantial performance gain compared with previous approaches to MDC using pairwise correlating transforms.