Cross-power spectrum phase for automated registration of multi/hyperspectral data-cubes for efficient information retrieval

A robust, feature-independent method of registration using the difference in phase between two images is applied in this work to the automated as well as fast registration of multi/hyperspectral spatial planes. Phase-difference is obtained using the cross spectrum of the two images of interest; properties of the cross spectrum allow its use in determining both rotation and translation compensation. In addition, dimension reduction of a sixty-band registered hyperspectral data-cube was implemented via principal component analysis. The eigenvectors generated during the diagonalization of the data´s covariance matrix were then used to reconstruct the cube´s original sixty spectral bands to various degrees of precision depending on the level of dimensional reduction previously performed