Matlab¶

Instructions¶

The test codes are built into the matlab files themselves, in the form of self-contained test and direct calculation functions near the end of each file. To run, simply call the function without any input arguments. The test functions also demonstrate appropriate usage, which is more straightforward than in C++. The input vectors can be modified to determine what input precision is appropriate for a particular regime.

Note that the programs are self-contained – they are not Matlab interfaces to the C++, but pure Matlab code. As such, the only setup step is to ensure the FINUFFT matlab interfaces are on the matlab path.

To use the FINUFFT Matlab interfaces, follow their documentation. Then, use Matlab’s “addpath” to ensure the FINUFFT programs (e.g. finufft2d3) will be found.

http://finufft.readthedocs.io/en/latest/index.html

Finally, make sure lgwt.m is in the same directory, or that its directory is also included in Matlab’s path.

Reconstruction Examples¶

In the recon directory, the autoquad functions will compute quadrature weights based on the coordinates sampled in k-space (see [1]). recon2d.m and recon3d.m use these weights in several different ways to reconstruct simulated data (the Shepp-Logan phantom). Note that recon3d is still under development, as autoquad3d can handle only small numbers of samples in a reasonable amount of time.

If using the recon examples from within the recon directory, also ensure that the directory one level above (containing all the sinc programs) is on Matlab’s path.