Install instructions for the 2D separable filter approximation code
                   with tensor decomposition 

+ REQUIRED libraries and toolboxes
- Teem/nrrd library [1], in particular the MATLAB version of the nrrd
  loader/writer (the files nrrdLoad and nrrdSave). Once compiled for your
  architecture, copy them in the nrrd/ subdirectory.
- Poblano toolbox [2]
- Tensor toolbox [3]

+ Filter banks
Copy the candidate filter bank, in text format with the filters represented 
as stacked square matrices, in the "data/" directory.
An example filter bank for the DRIVE dataset [4], has been included.

+ Run
Run separable_filter_approximation_td function. The resulting separable
filter bank will be stored in results directory along with the
reconstruction and reconstruction coefficients. The obtained separable
filter banks can then be used for the pixel classification task [5].

For any question or bug report, please feel free to contact me at:
bugra <dot> tekin <at> epfl <dot> ch