Variance-covariance/Principal components

Option on Image analysis window after loading Multigrids for the satellite image.

  • Basic statistics: create table with mean, std dev, min, and max for each band
  • Correlation matrix: show the correlation matrix for each pair of bands.
  • Variance-covariance matrix: show this matrix for each pair of bands.  The correlation matrix is computed from this matrix and is probably easier to interpret.
  • Eigen vectors and PC loadings: shows the relative magnitude of each eigenvector, and the loadings on each vector of each variable.
  • Principal component results: show the PC images.
  • Load PC images
  • Max PC bands: only this number of new grids will be created.  Get a senses of the number of pick by looking at the eigen vectors.  The 25 shown here only makes sense of hyperspectral imagery.
  • Min explanation to show (%): the higher principal components typically only show noise, so you can limit the images to display by specifing the minimium amount of the variation a component must explain.

Last revision 3/6/2017