Chapter 6
Irregular Variable-Length Codes for Joint Source and Channel Coding1
6.1 Introduction
As described in Chapter 2, Shannon’s source and channel coding separation theorem [133] states that, under certain idealized conditions, source and channel coding can be performed in isolation without any loss in performance. This motivated the design of the Vector Quantization (VQ) [170] based video transmission scheme of [273], in addition to the MPEG-4 [158] based scheme of [274]. Following video encoding in the transmitters of these schemes, the resultant bit sequences were protected using a serial concatenation [275] of Reversible Variable-Length Codes (RVLCs) [276] and Trellis-Coded Modulation (TCM) [277]. In the corresponding receivers, iterative A Posteriori Probability (APP) Soft-Input Soft-Output (SISO) TCM and RVLC decoding [278] was employed, and the resultant reconstructed video-encoded bit sequence was decoded. As described in Chapter 3, RVLC encoders typically fulfill a joint source and channel coding role. However, in the schemes of [273] and [274] the RVLC encoder was unable to achieve any source compression, but played a beneficial channel coding role, similar to that of the TCM encoder. This was because only a negligible amount of residual redundancy was apparent in the video-encoded bit sequences, owing to the high level of video compression achieved. The schemes of [273] and [274] may therefore be deemed to employ a separate source and channel coding philosophy. ...
