74 4 Statistical Pattern Recognition y' y X' X Fig. 4.23 Data points in feature space (assumed bivariate Gaussian) and a corresponding isocontour 10. Tabulate the data shown in Fig. 4.23 (download it from the http://extras. springer.com) and obtain the coordinates [in (x, y)] using an imaging program (Image] or Photoshop, or similar). Construct the covariance matrix (in the (x, y) coordinate system) from the data shown. Diagonalize the matrix (to get the coordinates in (x', y')) (Use [V, D] = eig in MatLab: the eigenvalues of A are obtained in D, and the columns of V are the eigenvectors of A)