Eigenanalysis of digital images in the Fourier domain: construction of prototypes for high-resolution human chromosomes

Eigenanalysis is a mathematical approach used to obtain characteristic roots and vectors from a matrix and is an important method for extracting information from digital images in microscopy. To use multiple images as input data, the images must be correctly aligned and the pattern common for each input image must not be geometrically distorted. Otherwise, the different geometries produce blurring. Alignment and removal of geometric distortion from digital images of random examples of 20 high resolution preparations of human karyotypes has been achieved by the eigenanalysis performed in the Fourier spectral domain on the phase and the amplitude of the Fourier transform. Our results show that eigenanalysis in the frequency & un provides a better resolution of features than does eigenanalysis in the spatial domain and makes it possible to construct statistically based prototypes.