Image Processing Reference
In-Depth Information
(a)
(b)
FIGURE 2.12
(a) Original LENA Image (b) aliased LENA Image.
2.4.2 I
MAGE
Q
UANTIZATION
Once the signal has been sampled (discretized), the amplitude of the sampled values
is quantized (digitized). A quantizer is a many-to-one mapping that maps a range of
input values into a single output value. There are two ways to quantize a signal:
scalar quantization (SQ), where each signal sample is individually quantized,
and vector quantization (VQ), where a block of signal samples is jointly quantized.
The concept of SQ for 1-D signals is shown in Figure 2.13. A typical
B
-bit scalar
quantizer for a signal with a dynamic range of x
min
to x
max
is de
ned as
2
B
Q
(
x
) ¼
r
i
if
d
i
1
<
x
<
d
i
i
¼
1, 2,
...
, L
where
L
¼
(
2
:
40
)
d
i
'
s are called decision levels and r
i
'
s are reconstruction levels.
The function Q
(
x
)
is shown in Figure 2.14.
2.4.2.1 Uniform Quantization
In uniform SQ, the quantization decision levels are of equal length
, referred to as
the quantization step size. The quantization step size is related to the signal dynamic
range
D
[
x
min
x
max
]
and number of quantization bits B by
f
(
n
)
f
(
x
)
f
Q
(
n
)
x
x
FIGURE 2.13
SQ for 1-D signals.
Search WWH ::
Custom Search