Measuring the textures in Image with yields the following results:
If the distances between the same textures are measured with the following result is obtained:
The results are confirmed with which gives:
That is, texture A and C are the most similar ones and textures A and B the least similar ones.
The final test of the metrics was to generate 100 triples of the kind that is shown in Image , and then to calculate the most and least similar textures. This was correlated to the ground truth and the following results were obtained
The human classification is performed in the same way as the automatic ones, that is to identify the most and least similar textures.
The probability of picking the most similar pair (or least similar at random is and the probability of picking both is .