It should be noted that the minimum bit rate required to transmit X with our approach is very dependent on the particular Huffman code for a given probability of occurence of 0's and 1's. As a result, for the cases, in which the Huffman code results in bit-planes with close probabilities of 0's and 1's, the transmitted bit rates are acceptable (although higher than the Slepian-Wolf bound); however, for the other cases the transmitted bit rates are very high.
It is also important to note that at a fixed probability of occurence of 0's (or 1's), the increase in the bit-rates, as p is increased, gets too high for p>0.10. The reason is, as also explained in the RESULTS, increase in probability of error causes an even greater increase in the probability of error of the M length Huffman coded codewords. This results in a difficulty in using our decoding approach for the first few bit-planes. Although using larger values of M might seem to give better results, it should be noted that larger M values is another factor that contributes to the increase in the error probabilities of the Huffman coded data. Hence, a trade-off needs to be made between the length of the Huffman codewords and the probability distributions of the Huffman coded bit-planes, and M=4 seems to be a good value to choose.
A better approach to the problem might be to use longer and better error-correcting codes, an example of which might be Reed-Solomon codes. Another solution might be to use a code other than a Huffman code as this is not to be the optimal method for getting equally likely 0's and 1's for the presented method.
ABSTRACT
INTRODUCTION
PRIOR WORK
BASIC SCHEME
METHODOLOGY
RESULTS
CONCLUSIONS
REFERENCES