Residual image coding using a SPIHT-based video coder

Abstract

This paper describes the design and implementation of a block-based set partitioning in hierarchical trees (SPIHT) coder which is capable of encoding residual images in low bit-rate video coding. Unlike the SPIHT coder which encodes an entire image frame, the new algorithm encodes the subdivided image regions (16times16 blocks) differentially based on an optimization bit allocation strategy. Instead of applying DWT to the whole image, it is applied separately on each partitioned block of the original image and, depending on the statistics of each image region, the number of wavelet coefficients to be retained is determined. Our experiments show that the proposed block-based SPIHT coder (BSC) performs better than SPIHT in terms of PSNR (objective criteria) & visual perception quality (subjective criteria)