Additive distortion modeling for unequal error protection of scalable multimedia content

Due to the improvement of compression technology, limited bandwidth channels can convey more data than before. However, since compressed streams are very sensitive to transmission errors, a joint optimization of the source-channel coding is required. Unequal error protection (UEP) is the most pragmatic way to do it, and exploits the embedded nature of bitstreams produced by scalable coders. However, depending on the number of substreams and protection levels, optimization of UEP can be extremely complex. A large simplification comes from considering the contribution to the global distortion of each substream as independent from the others. This paper explores this distortion additivity assumption for JPEG2000, and shows that though the assumption does not fully hold, no visible consequence on UEP performance is introduced, while reducing significantly the complexity of the optimization process.