Image Processing Reference
In-Depth Information
Q
(
x
)
r
L
r
3
r
2
r
1
x
d
0
d
1
d
2
d
3
d
L
-1
d
L
FIGURE 2.14
Quantization function Q
(
x
)
.
x
max
x
mi
L
¼
x
max
x
min
2
B
D ¼
(
2
:
41
)
The decision boundaries are related to the quantization step size
D
by
d
i
d
i
1
¼ D
1
i
L
(
2
:
42
)
The reconstruction levels are chosen to be halfway between the decision boundaries
in order to minimize the mean-squared error (MSE).
d
i
þ
d
i
1
2
r
i
¼
1
i
L
(
2
:
43
)
Example 2.4
Design a uniform 2 bit (4-level) quantizer for a signal with dynamic range between
0 and 1.
S
OLUTION
The decision boundaries and reconstruction levels are
d
0
¼
0,
d
1
¼
0
:
25,
d
2
¼
0
:
5,
d
3
¼
0
:
75,
d
4
¼
1
r
1
¼
0
:
125,
r
2
¼
0
:
375,
r
3
¼
0
:
625,
r
4
¼
0
:
875
2.4.2.2 Signal-to-Quantization Noise Ratio (SQNR)
Let
s assume
that the signal distribution (histogram) over a given quantizer decision interval
can be approximated by a uniform probability density function (PDF) as shown in
D
denote the quantization step size of a uniform quantizer and let
'
Search WWH ::
Custom Search