Cryptography Reference
In-Depth Information
the image. The chance of having small values are even higher for pixels that
are co-sited in the neighboring frames. Therefore a pixel could be predicted
from its neighboring pixels been coded. This is called Intra Predictive Coding.
Similarly, due to high temporal redundancy, a pixel or a block of pixels can
also be predicted using the pixels of the previous coded frames. This is called
Inter predictive coding. Predictive coding reduces the entropy of the signal
only and is lossless in principle. Further data reduction can be realized by
applying a coarse quantization to the predictive error signal and entropy cod-
ing techniques are used to the resulting signal. To avoid transmission error
propagation, a full value is transmitted periodically to provide an updated
reference.
Compressed
image
Compressed
image
f
n
f
n
+
+
e
n
e
n
e
n
e
n
Input
image
Input
image
quantiser
quantiser
_
_
de-quantiser
e
n
de-quantiser
e
n
e
n
f
^
f
^
+
+
f
f
n
f
n
predictor
predictor
+
+
(a) Predictive encoding
(a) Predictive encoding
Compressed
image
Compressed
image
e
n
e
n
f
n
f
n
Decompressed
image
Decompressed
image
de-quantiser
de-quantiser
e
n
e
n
f
^
f
^
f
predictor
predictor
(b) Predictive decoding
(b) Predictive decoding
Fig. 2.3.
Predictive encoder and decoder.
The principle of lossy predictive coding is illustrated in Fig. 2.3. It can be
seen that as each successive pixel f
n
enters the predicator, it generates the
anticipated value of that pixel based on past inputs,
f
n
= Pr(f
n−1
,f
n−2
, ..., f
n−i
).
(2.3)
The prediction error is given as,
−f
n
.
e
n
= f
n
(2.4)
The above error will be quantized as e
′
n
to further reduce the number
of bits used to describe this error signal. The quantization is a process that
