Principal components analysis (PCA) is one of the oldest and most important transformations of multivariate data analysis. The central idea is to generate linear combinations of the input data variables that are uncorrelated and have maximum variance. This reduces the dimensionality of the data while enhancing the features of interest.

In remote sensing this technique can be advantageously used to reduce the number of bands that are necessary for a certain analysis (i.e. classification) and so reduce computing costs while keeping as much as possible of the variability present in the data. Most GIS and remote sensing software packages in use today have implemented this function in some or another way. In practice, it is enough for an analyst to just press a virtual button to calculate te principal components of an image. This is comfortable but boring. It robs us of the fun of understanding the basic principles and…