Next: Appendix C: Proof of
Up: Bibliography
Previous: Appendix A: Proof of
First, we summarize the wavelet transform and the inverse wavelet transform[Chui1992] for designing a constant Q filterbank.
The integral wavelet transform for f(t) is defined by
|
(42) |
where a is the ``scale parameter,'' b is the ``shift parameter,''
with ,
and
is the conjugate of .
The integral basis function is
scale-transformed by parameter a and shifted by parameter b.
The selection of
allows much mathematical freedom; however, in general,
is determined to be an integrable function that satisfies the following ``admissibility condition'':
|
(43) |
where
is the Fourier transform of ,
It follows that
is a continuous function, so that the finiteness of
in Eq. (43) implies that
,
or equivalently,
.
If the above equation is satisfied,
is called a ``basic wavelet,'' and a unique inverse transform exists as follows:
|
(44) |
Since the analyzing wavelet
in Eq. (20) approximately satisfies the admissibility condition because
,
it can be considered that this analyzing wavelet is the basic wavelet.
Equations. (42)-(44) are a continuous wavelet transform.
The discrete wavelet transform corresponding to these equations is represented by
|
(45) |
|
(46) |
and
|
(47) |
where p and q are integer parameters.
Next: Appendix C: Proof of
Up: Bibliography
Previous: Appendix A: Proof of
Masashi Unoki
2000-11-07