Intelligent source coding

Where the error characteristics of the communications channel are known, it is possible to  bias the source encoding/decoding process deliberately to be tolerant of a certain type of error.

The images shown here are good examples of how intelligent source coding can pay off. The first image uses conventional image coding techniques designed for data channels that exhibit a random distribution of bit errors with time. When subject to a channel that produces short, focused bursts of errors, the decoded image is severely degraded. The second image has been passed through an encoder/decoder designed to be resilient to burst errors, and clearly demonstrates the marked improvement that can be achieved if the source coding is tailored to the channel error conditions - where known.

(Images kindly provided by Professor David Bull, Centre for Communications Research, University of Bristol.)