φt=2∑nhnφ2t−n
φ
t
2
n
h
n
φ
2
t
n
(1)
∫φtdt=∫2∑nhnφ2t−ndt
t
φ
t
t
2
n
h
n
φ
2
t
n
(2)
∫φtdt=2∑nhn∫φ2t−ndt
t
φ
t
2
n
h
n
t
φ
2
t
n
(3)
∫φtdt=2∑nhn∫φ2tdt
t
φ
t
2
n
h
n
t
φ
2
t
(4)
∫φtdt=2∑nhn∫φ2tdt
t
φ
t
2
n
h
n
t
φ
2
t
(5)
∫φtdt=2∑nhn12∫φtdt
t
φ
t
2
n
h
n
1
2
t
φ
t
(6)
1=12∑nhn
1
1
2
n
h
n
(7)
assuming
∫φtdt≠0
t
φ
t
0
. For a solution
φt
φ
t
to even exist, it is required that the coefficients
hn
h
n
add up to
2
2
.
∑nhn=2
n
h
n
2
(8)
φt=2∑nhnφ2t−n
φ
t
2
n
h
n
φ
2
t
n
(9)
ℱLHS=ℱRHS
ℱ
LHS
ℱ
RHS
ℱφt=ℱ2∑nhnφ2t−n
ℱ
φ
t
ℱ
2
n
h
n
φ
2
t
n
Φω=2∑nhnℱφ2t−n
Φ
ω
2
n
h
n
ℱ
φ
2
t
n
Φω=2∑nhn∫-∞∞φ2t−nⅇ-ⅈωtdt
Φ
ω
2
n
h
n
t
φ
2
t
n
ω
t
For
τ=2t−n
τ
2
t
n
,
Φω=2∑nhn12∫-∞∞φτⅇ-ⅈωτ+n2dτ
Φ
ω
2
n
h
n
1
2
τ
φ
τ
ω
τ
n
2
Φω=12∑nhnⅇ-ⅈω2n∫-∞∞φτⅇ-ⅈω2τdτ
Φ
ω
1
2
n
h
n
ω
2
n
τ
φ
τ
ω
2
τ
Φω=12∑nhnⅇ-ⅈω2nΦω2
Φ
ω
1
2
n
h
n
ω
2
n
Φ
ω
2
Φω=12Φω2∑nhnⅇ-ⅈω2n
Φ
ω
1
2
Φ
ω
2
n
h
n
ω
2
n
Note that
H
f
ω=DTFThn=∑nhnⅇ-ⅈωn
H
f
ω
DTFT
h
n
n
h
n
ω
n
is the discrete-time Fourier transform (DTFT) of the scaling
filter
hn
h
n
. We then have the Fourier form of the dilation
equation:
Φω=12
H
f
ω2Φω2
Φ
ω
1
2
H
f
ω
2
Φ
ω
2
(10)
When
ω=0
ω
0
is put into this equation, we get
Φ0=12
H
f
0Φ0
Φ
0
1
2
H
f
0
Φ
0
or
H
f
0=2
H
f
0
2
(11)
which is exactly the same equation as
∑nhn=2
n
h
n
2
that we already got.
From Equation 10, we can write
Φω2=12
H
f
ω4Φω4
Φ
ω
2
1
2
H
f
ω
4
Φ
ω
4
or
Φω=12
H
f
ω212
H
f
ω4Φω4
Φ
ω
1
2
H
f
ω
2
1
2
H
f
ω
4
Φ
ω
4
If we iterate, we get the infinite-product
formula for
Φω
Φ
ω
,
Φω=Φ0∏k=1∞12
H
f
ω2k
Φ
ω
Φ
0
k
1
1
2
H
f
ω
2
k
(12)
From this formula, we can see that the zeros of
Φω
Φ
ω
are determined by the zeros of
H
f
ω
H
f
ω
.
Given
hn
h
n
, what is the support of
φt
φ
t
?
Suppose
hn
h
n
is supported on
0≤n≤N−1
0
n
N
1
then from the dilation equation
φt=2∑n=0N−1hnφ2t−n
φ
t
2
n
0
N
1
h
n
φ
2
t
n
the support of
φt
φ
t
is found by matching the support of the left hand
side and the right hand side.
Suppose the support of
φt
φ
t
is
ab
a
b
. Then
-
φ2t
φ
2
t
has support
a2b2
a
2
b
2
-
φ2t−1
φ
2
t
1
has support
a+12b+12
a
1
2
b
1
2
-
⋮⋮
-
φ2t−k
φ
2
t
k
has support
a+k2b+k2
a
k
2
b
k
2
-
⋮⋮
-
φ2t−(N−1)
φ
2
t
N
1
has support
a+N−12b+N−12
a
N
1
2
b
N
1
2
The support of the LHS is
ab
a
b
by assumption. The support of the RHS is
a2b2+N−12
a
2
b
2
N
1
2
. Therefore, matching the endpoints we get
a=a2
a
a
2
b=b2+N−12
b
b
2
N
1
2
which gives
a=0
a
0
b=N−1
b
N
1
The support of
φt
φ
t
is
0N−1
0
N
1
.
hn has finite support ⇒φt of finite support
h
n
has finite support
φ
t
of finite support
That means that there are
finite degrees
of freedom in design of finitely supported scaling functions.
Note:
limj→∞12j∑lφl2j=∫φtdt
j
1
2
j
l
φ
l
2
j
t
φ
t
Let us define
∀j,j≥1:
S
j
≔12j∑lφl2j
j
j
1
≔
S
j
1
2
j
l
φ
l
2
j
Begin with the dilation equation
φt=2∑nhnφ2t−n
φ
t
2
n
h
n
φ
2
t
n
and substitute
l2j
l
2
j
for tt.
φl2j=2∑nhnφl2j−1−n
φ
l
2
j
2
n
h
n
φ
l
2
j
1
n
with
j≥1
j
1
. Now sum over ll:
∑lφl2j=∑l2∑nhnφl2j−1−n
l
φ
l
2
j
l
2
n
h
n
φ
l
2
j
1
n
∑lφl2j=2∑nhn∑lφl2j−1−n
l
φ
l
2
j
2
n
h
n
l
φ
l
2
j
1
n
∑lφl2j=2∑nhn∑lφl2j−1
l
φ
l
2
j
2
n
h
n
l
φ
l
2
j
1
∑lφl2j=2∑lφl2j−1
l
φ
l
2
j
2
l
φ
l
2
j
1
Divide both sides by
2j
2
j
:
12j∑lφl2j=12j−1∑lφl2j−1
1
2
j
l
φ
l
2
j
1
2
j
1
l
φ
l
2
j
1
S
j
=
S
j
−
1
S
j
S
j
−
1
For
j≥1
j
1
, all the
S
j
S
j
are the same!
limj→∞
S
j
=
S
0
j
S
j
S
0
Therefore
∫φtdt=∑kφk
t
φ
t
k
φ
k
(13)
If
φt
φ
t
is continuous, then it can also be shown that
∫φtdt=∑kφ
t
o
+k
t
φ
t
k
φ
t
o
k
(14)
for any real
t
o
t
o
.