An unequal error protection scheme for SPIHT image transmission with prioritised retransmissions and de-noising

Abstract

This paper proposes a robust resilient transmission scheme for images that have been compressed using the Set Partitioning In Hierarchical Trees (SPIHT) algorithm. The scheme employs Unequal Error Protection (UEP), prioritised Automatic Repeat reQuest (ARQ) and two-dimensional adaptive Wiener filtering. In order to apply UEP, the SPIHT coded bit-stream is partitioned into bit-classes of unequal importance. UEP is then applied at two levels. Firstly, Rate Compatible Punctured Convolutional (RCPC) codes are used to encode the bit-classes with different code-rates hence providing them with different levels of protection. Secondly, prioritized retransmissions are used such that the more important bit-classes are allocated a greater number of retransmissions than the less important ones. Equal gain combining is also performed to further improve the performance. The proposed U E P s c h e m e provides an average gain of 3 dB in P e a k S i g n a l to Noise Ratio (PSNR) over a conventional Equal Error Protection (EEP) in the range 0dB≤Eb/No≤6dB with 16-QAM and an average gain of 3.5 dB in PSNR in the range 0dB≤Eb/No≤3dB with BPSK modulation. Moreover, de-noising with a two dimensional Wiener filter provides an additional gain of 0.76 to 0.90 dB.