Comparison of PCA, NNMF, and ICA
See this discussion for doing Independent Component Analysis (ICA) in Mathematica : "Independent component analysis for multidimensional signals".
This blog post shows an extension of the comparison in this discussion with ICA : "Comparison of PCA, NNMF, and ICA over image de-noising".
Image highlights of that comparison follow.
Original, noised, and de-noised images
Comparison using classifiers
We can further compare the de-noising results by building digit classifiers and running them over the de-noised images.
We can see that with ICA we get better results than with PCA/SVD, probably not as good as NNMF, but very close.