quantifies the level of randomness in the region and has its highest value when all c^ are equal. This is the case where all intensity transitions are equally likely. Although this case is often not observed in practice, it illustrates the type of information that this metric conveys. The entropy metric may be used to distinguish tissue with somewhat structured texture from tissue with less structure. The first yields a relatively lower entropy than the second.

Other texture metrics can be computed by using marginal distributions derived from the co-occurrence matrix:

The means and standard deviations of these marginal distributions are represented by , , <jx and ay. In addition, the sums of probabilities that relate to specified intensity sums or differences are defined by cx+y«= E cij * = 0, 1, 2,..., 2K - 2 (43)

and cx - y (*) = E cij * = 0, 1, 2,..., K - L (44) |i - j| = *

Some texture metrics based on these distributions are:

Correlation fc

0 0

Post a comment