(*) Modified Huffman coding with a fixed symbol table size of 256
symbols was used.
The poorer efficiency with the olivine data is a combination of the
hight frequency on 0 values and a
result of the symbol table being too small. (imgCIF / CBF will allow
the number of symbols stored to be a variable size, but we still need
an algortihm which can determine an optimum or close to optimum number
of symbols.)
We can compare the efficiency of the Huffman algorithm to the theoretical
maximum compression based on the zeroth order entropy.
ztyl01.pmi
Predictors
1
2
3
4
Entropy Compression
3.25
3.45
3.39
3.44
Huffman Comprresion (*)
3.22
3.41
3.36
3.40
Huffman Efficiency
99.2%
98.9%
99.0%
98.9%
olivine.image
Predictors
1
2
3
4
Entropy Compression
2.18
2.28
2.22
2.25
Huffman Compression (*)
2.02
2.12
2.05
2.08
Huffman Efficiency
92.4%
93.0%
92.6%
92.7%