Database Reference
In-Depth Information
{
[
,
]
}
=
[
+
−
, −
+
−
]
g
i
k
f
i
cos
ʸ
k
sin
ʸ
a
i
sin
ʸ
k
cos
ʸ
b
(6.15)
0
0
0
0
where index 0
1 for an image of size
M
row-pixels and
N
column-pixels. Following the property of Fourier transformation [
173
], we can
obtain the Fourier transformation of
≤
i
≤
M
−
1 and 0
≤
k
≤
N
−
{
g
[
i
,
k
]
}
that is related to the Fourier transform
of
{
f
[
i
,
k
]
}
by:
F
m
cos
ʸ
0
(6.16)
nM
N
mN
M
e
−
j
2ˀ
(
M
+
n
N
)
×
m
G
[
m
,
n
]=
ʸ
0
+
sin
ʸ
0
,−
sin
ʸ
0
+
n
cos
[
,
]
with the corresponding magnitude of
G
m
n
which is equal to:
F
m
cos
ʸ
0
nM
N
mN
M
ʸ
0
+
sin
ʸ
0
,−
sin
ʸ
0
+
n
cos
(6.17)
=
When
M
N
, we have the following relationship:
|
[
,
]
|
=
|
[
+
,−
+
]
|
G
m
n
F
m
cos
ʸ
n
sin
ʸ
m
sin
ʸ
n
cos
ʸ
(6.18)
0
0
0
0
Let
|
G
[
m
,
n
]
|
=
M
G
[
m
,
n
]
and
|
F
[
m
,
n
]
|
=
M
F
[
m
,
n
]
, and then the relationship
shown in Eq. (
6.18
) can be rewritten as:
M
G
[
m
,
n
]=
M
F
[
m
cos
ʸ
0
+
n
sin
ʸ
0
,−
m
sin
ʸ
0
+
n
cos
ʸ
0
]
(6.19)
From the relationship in Eq. (
6.19
), the translation parameters disappear, leaving
only rotation parameters. The estimation of the parameters for rotation starts with
changing the magnitude of the Fourier transform from rectangular coordinates
[
m
,
n
]
to polar coordinates
[
r
,
ʸ
]
by substituting index
m
=
r
cos
ʸ
and index
n
=
r
sin
ʸ
into
Eq. (
6.5
). The Fourier transform in polar coordinates can be obtained by:
M
−
1
i
=
0
N
−
1
k
=
0
g
[
i
,
k
]
e
−
j
2ˀ
(
r
cos ʸ
1
MN
)
M
+
N
r
sin
ʸ
G
[
r
,
ʸ
]=
(6.20)
Similarly, the Fourier transform in polar coordinates of
f
[
i
cos
ʸ
0
+
k
sin
ʸ
0
−
a
, −
i
sin
ʸ
0
+
k
cos
ʸ
0
−
b
]
can be obtained by:
1
MN
e
−
j
2ˀ
(
(
m
cos ʸ
0
+
n
N
sin
ʸ
0
)
M
+
(
−
m
M
sin
ʸ
0
+
ncos
ʸ
0
)
u
v
N
)
u
∑
v
f
[
u
,
v
]
(6.21)
Next, substituting
m
=
rcos
ʸ
and
n
=
rsin
ʸ
into Eq. (
6.21
) produces the Fourier
transform in polar coordinates:
f
[
u
,
v
]
e
−
j
2ˀ
(
u
M
+
v
N
)
=
u
∑
v
r
cos
(
ʸ
−
ʸ
0
)
r
sin
(
ʸ
−
ʸ
0
)
F
[
r
,
ʸ
−
ʸ
0
]
(6.22)
MN
Search WWH ::
Custom Search