PhD thesis

AuthorRüdiger Knörig
TitleMultiple Description Coding mittels kaskadierter korrelierender Transformationen
AbstractThis work covers a new multiple description transform coding approach which, seen as a joint source-channel-coder, provides stepless trade-offs between the source coding goal of a high compression ratio and the channel coding goal of a high robustness against channel errors.
The proposed solution consists of two main parts. The first one is a new kind of transformation, based on the idea of Goyal et al in [Goy01b] to cascade the elementary 2x2 correlation transform for gaining transforms of higher order.
The second part is a MSE-optimal estimator which uses this variable amount of correlation to estimate a reconstruction of the input signal from a subset of transform coefficients. This would be necessary in case of transmission errors resulting in coefficient losses.
With this approach one can select fine-tuned trade-offs between efficient compression and channel error robustness depending on the actual channel situation. Therefore this approach will be most useful with channels strongly varying regarding their transmission quality, like wireless or internet transmissions.
Key wordsjoint source-channel-coder, MSE, estimator, DST transform, transmission, multiple description coding, Quellencodierung, Kanalcodierung, CCT, Transformation, JPEG, Goyal
NoteNr. 24