Image Processing Reference
In-Depth Information
1
0.8
0.6
0.4
0.2
0
0
50
100
150
200
250
Code value
f
FIGURE 2.45
CDF of LENA image.
2.8.3 H
ISTOGRAM
E
QUALIZATION
Assume that r is a continuous random variable taking values between 0 and 1 with a
PDF p
r
(
r
)
. Let T
(
r
)
be a single-valued and monotonically increasing function of its
argument such that
0
T
(
r
)
1
for
0
r
1
(
2
:
128
)
Then, according to the theory of functions of one random variable, the random
variable s
¼
T
(
r
)
has a PDF given by
p
r
(
r
)
d
T
p
s
(
s
) ¼
d
r
j
r
¼
T
1
(
2
:
129
)
(
s
)
Now if we want the random variable s to be uniformly distributed over the interval
[
01
]
, then we set p
s
(
s
) ¼
1. Substituting this into Equation 2.129 yields
p
r
(
r
)
d
T
d
r
1
¼
(
2
:
130
)
or
d
T
d
r
¼
p
r
(
r
)
(
2
:
131
)
Integrating both sides of Equation 2.131, we have
ð
r
s
¼
T
(
r
) ¼
p
r
(
x
)d
x
(
2
:
132
)
0
Search WWH ::
Custom Search