Enhanced partial-SPIHT for lossless and lossy image compression

This paper proposes an enhanced partial SPIHT (P-SPIHT) for lossless and lossy image compression. P-SPIHT uses three coding modes for each bit plane based of the probability of the significant coefficients (PI) within each bit plane as proposed by Abu-Hajar and Sankar (see ICASSP 2002, Orlando, Florida, vol.4, p.3497-3500, 2002). In this paper, P-SPIHT is extended to support both lossy and lossless compression. Also it sorts the coded data into three categories; sign bits (SB), tree bits (TB) and magnitude bits (MB). P-SPIHT sorts TB and MB into insignificant tree bits (ITB), significant tree bits (STB), insignificant magnitude bits (IMB) and significant magnitude bits (SMB). Sorting the data enhances the compression of the arithmetic coder, as each category uses its own frequency model within the coder. Sorting the data improves the compression of the arithmetic coder by a factor of two. Experimental results show that the compression of P-SPIHT is superior to SPIHT for all the tested images and it is comparable JPEG2000 especially in the lossless mode.