Applications of Emergent Phenomena - Systematic Design for Emergence in Cellular Nonlinear Networks - page 145

of the defining parameters (such as the ID , the number of iterations T , the size C
of the codebook, and the block size w ).
The operating principle of the encoder and decoder employing the CA-
generated codebooks are presented in Figs. 8.12 and 8.13, respectively.
The encoding process is simple: The input image is split into a number of bit-
planes. Experiments with many natural images indicate that the most significant
bitplane ( m = 1) has the lowest content in high frequencies, while the least signifi-
cant bitplanes can be regarded as almost random. This observation led to the
following simplification in the encoding algorithm:

x The first m most significant bitplanes are encoded as indicated next.

x The remaining least significant bitplanes (called ignored bitplanes in

the next) are replaced with bitplanes generated pseudo-randomly using
a different key per bitplane (e.g. using a specific CA with its ID as a
compact key, see Fig. 8.11). It is assumed that the keys are known in
the decoder, so that ignored bitplanes identical to the ones encoded

could be generated.

Fig. 8.12. The principle of encoding (compression) using CA-generated codebooks

Each of m most significant bitplanes is divided into bit blocks of wxw bits
size each. For each such block B a best match codeword is found within the C

J

CA-generated codebook. It minimizes the Hamming distance (the

number of different bits between the actual block B and the selected codeword
J

d B C

J J

1

K is the index of the best match, being sent as the
compressed version of the initial block. Of course, since there is a limited number
of codewords, some errors will result from the approximation

). In the above J

^

1, C

`

C

. C #

J

B