MeasTex: Comparisons of Algorithms
The MeasTex framework computes a scalar measure of an algorithm's
classification performance on a suite of texture problems. Given these
quantitative measures, it is possible to compare different algorithms.
We have summarised the measures for the following comparisons:
-
Major paradigms
-
GLCM vs Markov vs Gabor
-
Variations of Gabor
-
Wide vs Narrow Gaussian envelopes
-
Varying number of frequencies measured
-
Variations of Markov
-
Varying Markov mask
-
Variations of GLCM
-
Varying distance parameter
-
Ohanian and Dubes Comparisons
-
Comparisons of algorithms similar to Ohanian and Dubes
Note: The scores quoted in these results were
obtained using the performance
measure described elsewhere. A score of 1 indicates that the
algorithm got the task completely right whereas a score of 0 indicates
that the algorithm was completely wrong. Because the measure combines
confidence and correctness, a range of results between these extremes
is possible.