Watch out with your eigenvalue analysis - your covariance matrix is limited to rank 500 because you're only using 500 images. Furthermore, the smaller eigenvalues will have eigenvectors with high variance due to the random image sample (ie. "noisy"). If the singular values start to fall off within ~1-3x of your sample size then that's indication you need larger sample - eg. random matrix spectra will still exhibit rolloff in spectra.
Thanks. The sample size was limited for memory restrictions. I obtained similar results in eigenvalues for different random selections of 500 images, but no doubt a covariance matrix based on more images might tell us more.