ICA – Independent Components Analysis

Independent Components Analysis – ICA – looks for statistical differences in the multispectral data and separates the data into statistically independent sets. At times, this can reveal even more difficult-to-see text than PCA can reveal.

ICA has two inputs, both of which are the same multiband image. The independent components are calculated from the region of interest specified in the “roi” image and the resulting ICA transform is applied to the “input” image. The two images must have the same number of bands, but they can be different sizes. There are three possible nonlinearities to apply. The “logcosh” is the default. The three different nonlinearities can give slightly different results that might reveal different characters. One caution in using ICA, the larger the “roi” region, the more spectral bands that are included in the input, and the larger the number of iterations, the longer the computation time will be.

The number of output bands is equal to the number of input bands. The output bands are ordered in decreasing spatial coherence. As a result, the low order bands have images with higher spatial correlation. Since text has high spatial correlation, the lower order bands tend to reveal text, both erased and visible.

In the “bandsOut” window, one can specify the number of output bands desired, which can be less than the number of input bands. If M output bands are specified, then bands 0 to M-1 will be output.