FIR wavelet filters are mainly used in conventional wavelet-based image cod
ing. However, FIR filters, except for the Haar wavelet, cannot satisfy both
the exactly linear phase and orthonormality conditions. For example, the w
ell-known Daubechies-9/7 wavelet is biorthogonal. In this paper, an effecti
ve implementation of the all-pass-based wavelet filters is presented for wa
velet-based image coding. Since all-pass filters are IIR, both the exactly
linear phase and orthonormality conditions can be simultaneously satisfied,
so that better compression performance can be expected, Also, IIR filters
can be realized with a lower computational complexity than can FIR filters.
Finally, the proposed IIR wavelet filters are evaluated by using SPIHT to
compress the practical images, and the influence of the all-pass filters an
d delay elements on the compression performance is investigated. It is show
n through the experimental results that the IIR wavelet filters proposed in
this paper have a lower computational complexity than the Daubechies-9/7 w
avelet, with a comparable compression performance. (C) 2001 Scripta Technic
a, Electron Comm Jpn Pt 3, 85(2): 13-21, 2002.