Error Detection in SPIHT Coded Images

Abstract

In this paper we propose a novel approach based on slice of information detection in order to improve the error detection efficiency and accuracy of the set partitioning in hierarchical trees (SPIHT) algorithm. If decoder detects error data more (or less) than the block number which is set in the slice head, it breaks and marks that error. After the whole image is decoded, the decoder starts to detect correlation according to inner-block correlation (IC) and interact-block correlation (ICB). These detection thresholds (IC & ICB) are updated by measuring the mean average over the error block's neighbor to locate the erroneous block. The erroneous block (whose parameter (IC or ICB) is greater than the threshold) is dealt with error concealment according to error concealment neighbor way. The simulation results show that the algorithm is simple, and can effectively find the errors. Comparing with the method without error detection and concealment, the decoder PSNR of reconstructed image has improved 0.3-1.1 dB.