Image Processing Reference
In-Depth Information
f
UM
(
n
,
m
)
-
d
(
n
,
m
)
+
Low-pass
filter
β
f
(
n
,
m
)
f
(
n
,
m
)
+
+
FIGURE 2.42
Unsharp masking.
the equivalent FIR
filter starting from the original image f
(
n, m
)
to the output
unsharp masked image f
UM
(
n, m
)
is given by
h
UM
(
n, m
) ¼ (
1
þb)d(
n, m
) b
h
LP
(
n, m
)
(
2
:
122
)
For example, if the LPF h
LP
(
n, m
)
is a simple 5
5 moving average
filter, then
2
4
3
5
b b b b b
b b b b b
b b
1
25
h
UM
(
n, m
) ¼
b b b
b b b b b
b b b b b
24
þ
25
(
2
:
123
)
25
The following two unsharp masked
filters are designed with
b ¼
11
¼
2
:
2727 and
b ¼
25
4
¼
:
6
25.
2
4
3
5
1
1
1
1
1
1
1
1
1
1
1
11
h
UM
(
n, m
)j
b¼
2
:
2727
¼
1
1 5
1
1
(
2
:
124
)
1
1
1
1
1
1
1
1
1
1
and
2
3
1
1
1
1
1
4
1
1
1
1
1
5
1
4
h
UM
(
n, m
)j
b¼
6
:
25
¼
1
1 8
1
1
(
2
:
125
)
1
1
1
1
1
1
1
1
1
1
In general, higher values of
b
result in more image sharpening. Similarly, in the
above
filters, a smaller center tap implies more sharpening since it increases the
relative contribution of the negative taps on the
filtered output.
2.8.2 I
MAGE
H
ISTOGRAM
Consider an M
M, 8 bit image f
(
n, m
)
, for each code
value f is the number of times the code value f appears in the image. The PDF of
image f
(
n, m
)
. The image histogram N(f
(
f
)
is the normalized histogram de
ned by
N
(
f
)
M
2
p
(
f
) ¼
f
¼
...
(
:
)
0, 1,
, 255
2
126
Search WWH ::
Custom Search