krober10nd / HamJacobi

Solves a Hamilton-Jacobi PDE fast (in seconds) to gradient limit a scalar field defined in 2D or 3D.

HamJacobi

Solves a Hamilton-Jacobi PDE fast (in seconds) to gradient limit a scalar field defined in 2D or 3D. The input to the solver is packed in column-major order with z being the slowest varying dimension.

                       

Compile

This code is designed to be mex'ed using https://github.com/audiofilter/mex-it. From MATLAB enter the following command:

mex CXXFLAGS="$CXXFLAGS -std=c++11" FastHJ.cpp

Note: you MAY have to start MATLAB from the terminal (on Linux-like OS) like so:

LD_PRELOAD=/usr/lib/x86_64-linux-gnu/libstdc++.so.6 matlab

...but first try to just use MATLAB without that.

Usage

Operate this code from MATLAB by changing the appropriate parts of the code below.

dims = [nrows ncols nz]; % note: nz MUST be 1 for 2D fields.  
elen = % size of grid cell 
dfdx = % decimal fraction representing smoothness
itmax = % maximum number of iterations to perform 
field = 2D array reshaped in column-major order 

smoothed_field = FastHJ( int32(dims), elen, dfdx, int32(itmax), field);

Column-major in MATLAB is readily achieved by using the reshape command like so:

[nrows,ncols,nz]=size(matrix);
vec = matrix(:); 

...and back to the original matrix format like:

matrix = reshape(vec,[nrows,ncols,nz]);