
i
i
i
i
i
i
i
i
436 19 Computational Information Theory
s
C={ c(1),c(2),...,c(N)
}
ŝ
z
Source
Encoder
Source
Decoder
Channel
Decoder
Channel
Encoder
x y ŷ
Figure 19.5 Channel denoising using Cloud sources.
approach presented in Chapter 17. Because channel noise is random, we can use the
existing correlation among pixels or coefficients to suppress additive channel noise.
Our experimental results show that the received quality is even better than that of
the state-of-the-art digital systems. However, this scheme has revealed limitations
because it only uses the correlation within input source. If channel SNR is too low
(i.e., channel noise is too large), all received data ...