Random Sampling of the Above Image Using a 9×9 Window

(1) Apply a 9×9 window to randomly sample the above image, extracting a total of 200 sub-images; (2) Convert each sub-image into an 81-dimensional row vector by concatenating all columns; (3) Perform KL transformation on all 200 row vectors, compute the eigenvectors and eigenvalues of the corresponding covariance matrix, and arrange eigenvalues and corresponding eigenvectors in descending order; (4) Select the eigenvectors corresponding to the top 40 largest eigenvalues as principal components, project the original image blocks onto these 40 eigenvectors, and use the obtained projection coefficients as the feature vector for each sub-block; (5) Compute the feature vectors for all sub-blocks.