Code covered by the BSD License  

Highlights from
Matlab code for my Graduate Thesis

from Matlab code for my Graduate Thesis by Troy
Numerically solves the diffusion equations as it pertains to medical imaging.

cnfft2D(varargin)
function w = cnfft2D(varargin)
% crankNicolson: uses Crank-Nicolson algorithm to approximate the solution
%   to the parabolic PDE (in 2-D):
%       u_{t}(x,t) - alpha^2 u_{xx}(x,t) = 0, 0<x<L, 0<t<T
%   subject to the boundary conditions
%       u(0,t) = u(L,t) = 0, 0<t<T
%   and the initial conditions
%       u(x,0) = f(x), 0<=x<=L
%
% arguments:
%   L (vector 1by2) - upper bound of spatial (x) variable
%       (Default L = [1,1])
%   T (scalar) - upper bound of time (t) variable
%       (Default T = 1)
%   alpha (scalar) - square root of coefficient of u_{xx} term
%       (Default alpha = 1)
%   m (vector 1by2) - number of discrete spatial intervals
%       (Default m = [100,100])
%   n (scalar) - number of discrete time intervals
%       (Default n = 100)
%
%   w (m+1 x m+1 x n) - approximation to u(x,t) at discrete space/time positions
%

% author: Troy J. Winkstern
% email: tjw8191@rit.edu
% date: 30 Jan 2011

% parse input arguments
[L,T,alpha,m,n] = parseInputs(varargin{:});
L_x = L(1);
L_y = L(2);
m_x = m(1);
m_y = m(2);

% initialize h, k, lambda, and w
h_x = L_x/m_x;
h_y = L_y/m_y;
k = T./n;
lambda_x = (alpha.^2).*k./(h_x.^2);
lambda_y = (alpha.^2).*k./(h_y.^2);
w = zeros(m_y+1,m_x+1,n);

% initialize w (change this to sin(pix1/l)*sin(pix2/l)
for i=2:m_x
    for j=2:m_y
        w(j,i,1) = sin(pi.*h_x.*(i-1)./L_x).*sin(pi.*h_y.*(j-1)./L_y);
    end
end


for t=1:(n-1)
    y = zeros(m_y+1,m_x+1);
    
    %Computes second column of the matrix
    y(2,2) = (1-lambda_x.*lambda_y)*w(2,2,t) + ...
        (lambda_y/2)*w(3,2,t)+ ...
        (lambda_x/2)*w(2,3,t);
    
    for j = 3:(m_y-1)
        y(j,2) = (lambda_y/2)*w(j-1,2,t) + (1-lambda_x.*lambda_y)*w(j,2,t) + ...
            (lambda_y/2)*w(j+1,2,t)+ ...
            (lambda_x/2)*w(j,3,t);
    end
    
    y(m_y,2) = (lambda_y/2)*w(m_y-1,2,t) + (1-lambda_x.*lambda_y)*w(m_y,2,t) + ...
        (lambda_x/2)*w(m_y,3,t);
    
    %Computes the inner parts of the matrix
    
    for i = 3:(m_x-1)
        y(2,i) = (1-lambda_x.*lambda_y)*w(2,i,t) + ...
            (lambda_y/2)*w(3,i,t)+ (lambda_x/2)*w(2,i-1,t) + ...
            (lambda_x/2)*w(2,i+1,t);
        for j = 3:(m_y-1)
            y(j,i) = (lambda_y/2)*w(j-1,i,t) + (1-lambda_x.*lambda_y)*w(j,i,t) + ...
                (lambda_y/2)*w(j+1,i,t)+ (lambda_x/2)*w(j,i-1,t) + ...
                (lambda_x/2)*w(j,i+1,t);
        end
        y(m_y,i) = (lambda_y/2)*w(m_y-1,i,t) + (1-lambda_x.*lambda_y)*w(m_y,i,t) + ...
            (lambda_x/2)*w(m_y,i-1,t) + ...
            (lambda_x/2)*w(m_y,i+1,t);
    end
    
    %Computes the m_x column of the matrix
    
    y(2,m_x) = (1-lambda_x.*lambda_y)*w(2,m_x,t) + ...
        (lambda_y/2)*w(3,m_x,t)+ (lambda_x/2)*w(2,m_x-1,t);
    
    for j = 3:(m_y-1)
        y(j,m_x) = (lambda_y/2)*w(j-1,m_x,t) + (1-lambda_x.*lambda_y)*w(j,m_x,t) + ...
            (lambda_y/2)*w(j+1,m_x,t)+ (lambda_x/2)*w(j,m_x-1,t);
    end
    
    y(m_y,m_x) = (lambda_y/2)*w(m_y-1,m_x,t) + (1-lambda_x.*lambda_y)*w(m_y,m_x,t) + ...
        (lambda_x/2)*w(m_y,m_x-1,t);
    
   
   % went out to length 2m
   X = sqrt(2./m_y).*imag(fft(y, 2*m_y,1));
   X = X(1:m_y+1,:);
   
   Y = sqrt(2./m_x).*imag(fft(X, 2*m_x,2));
   Y = Y(:,1:m_x+1);
   
   % angles divided by m
   a_x = (pi./m_x).*(0:m_x);
   a_y = (pi./m_y).*(0:m_y).';
   
   % changed lambda to accomodate for 2-D
   c = 1 + (lambda_x.*lambda_y).*(1-cos(a_y))*(1-cos(a_x));

   % went out to length 2m
   Z = sqrt(2./m_y).*imag(fft(Y./c, 2*m_y,1));
   Z = Z(1:m_y+1,:);
   
   R = sqrt(2./m_x).*imag(fft(Z, 2*m_x,2));
   w(2:m_y,2:m_x,t+1) = R(2:m_y,2:m_x);
end

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
% subfunction parseInputs
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
function [L,T,alpha,m,n] = parseInputs(varargin)

% check number of input arguments
nargs = length(varargin);
error(nargchk(0,5,nargs));

% get/set n
if nargs<5
    n = [];
else
    n = varargin{5};
end
if isempty(n)
    n = 100;
end

% check n has valid value
if (numel(n)>1) || (n<1) || ~isequal(round(n),n)
    error('n must be a positive integer.');
end

% get/set m
if nargs<4
    m = [];
else
    m = varargin{4};
end
if isempty(m)
    m = [100,100];
end

% check m has valid value
if (numel(m)~=2) || any(m<1) || any(~isequal(round(m),m))
    error('m must be a vector of positive integers.');
end

% get/set alpha
if nargs<3
    alpha = [];
else
    alpha = varargin{3};
end
if isempty(alpha)
    alpha = 1;
end

% check alpha has valid value
if numel(alpha)>1
    error('alpha must be a real scalar.');
end

% get/set T
if nargs<2
    T = [];
else
    T = varargin{2};
end
if isempty(T)
    T = 1;
end

% check T has valid value
if (numel(T)>1) || (T<=0)
    error('T must be a positive real scalar.');
end

% get/set L
if nargs<1
    L = [];
else
    L = varargin{1};
end
if isempty(L)
    L = [1,1];
end

% check L has valid value
if (numel(L)~=2) || any(L<=0)
    error('L must be a positive vector with 2 elements.');
end

% warning if L not integer
if any(~isequal(round(L),L))
    warning('L should be an vector of integers so that the inital condition takes on the value of 0 for t = 0 and t = T.');
end

Contact us