* 2010-03-22 *

This is a movie of a 224x224 covariance matrix corresponding

to the AVIRIS data, as estimated by SMT with K rotations,

with K ranging from 1 to 389. That corresponds to r = (389/224) = 1.74 rotations per coordinate.

The first frame is that actual sample covariance matrix,

and the following frames show how the matrix goes from diagonal to blockdiagonal, to a dense matrix.

One iteresting aspect is how the a single rotation can suddenly "populate" a large chunk of the

covariance matrix as (seemingly) two blocks are merged into a larger block.