Permuted smoothed descriptions and refinement coding for images

摘要

We consider the problem of transmitting compressed still imagesnover lossy channels. In particular, we examine the situation where thendata stream is partitioned into two independent channels, as is oftennconsidered in the multiple descriptions approach to image compression.nWe introduce a coder design called smoothed descriptions, which matchesna data partitioning method utilized at the encoder to the errornconcealment technique employed at the decoder. This approach has thenadvantage of inserting minimal overhead into the transmitted datanstreams, so that system performance is undiminished when there are nonpacket losses over the channel. We show that, by using a combination ofnDC averaging and maximal smoothing to conceal errors, performancencomparable to or better than multiple descriptions can be achieved fornpacket loss rates up to 5%. By adding a feedback loop that requestsnretransmission of important data, we also demonstrate the need tonexploit latency whenever possible