In todays sourcecoding technologies for audio- and videocompression an undistorted transmission between source and drain is assumed for the decoder to be able to reconstruct the encoded material without errors. The normally used codes of variable length (VLC) like the Huffman code or arithmetic coding are hereby very sentiv to transmission errors because they are coded in dependency of their predecessors. The distortion of a symbol thus does not affect only the symbol itself but also all following ones. As protection against transmission errors additional redundancy information is added by a channel coder at sender side which allows the receiver to correct distorted bits within limits.
Todays transmission systems assume each bit to be of equal importants. The channel encoding takes place thus independently of the transmitted contents. This is not true for video data because a predictions is used to predict the current frame out of already transmitted ones. Distortions in such reference frames thus propagate through all depended frames. It is therefore meaningfull to protect reference frames higher than other ones.
In our approach the video stream is devided into one or more single streams which are independently protected and transmitted. Different techniques for stream partitioning and uneaqual error protection such as hierachical modulation or usages of different channel coding rates for each stream are investigated.
