SVD has been part of my work for a long time now. For those who havent used it, the SVD or Singular Value Decomposition is an extremly powerful technique. Its the core routine for many applications, from dimensionality reduction to filtering to graph analysis to signal processing to pattern recognition and much much more.
Simply put SVD helps to do what your brain automatically did in recognizing the rose in the pieces of wood in the above picture.
I first came across the need for SVD when applyting the technique to compute Principal component analysis on manufacturing blades to measure manufacturing variability.
A few weeks back, stumbled upon this video demonstration of how SVD works on YouTube.
Watch “1976 Matrix Singular Value Decomposition Film”
If you are looking for SVD fortran code similar to matlab functionality then follow this link to SVD in fortranwiki