new file mode 100644
--- /dev/null
+++ b/fftconv2.m
@@ -0,0 +1,109 @@
+## FFTCONV2 Convolve 2 dimensional signals using the FFT.
+##
+## usage: fftconv2(a, b[, shape])
+##        fftconv2(v1, v2, a, shape)
+##
+## This method is faster but less accurate for large a,b.  It
+## also uses more memory. A small complex component will be 
+## introduced even if both a and b are real.
+##
+## see also: conv2
+
+## Author: Stefan van der Walt <stefan@sun.ac.za>, 2004
+
+function X = fftconv2(varargin)
+    if (nargin < 2)
+	usage("fftconv2(a,b[,shape]) or fftconv2(v1, v2, a, shape)")
+    endif
+
+    shape = "full";
+    rowcolumn = 0;
+    
+    if ((nargin > 2) && ismatrix(varargin{3}))
+	## usage: fftconv2(v1, v2, a[, shape])
+
+	rowcolumn = 1;
+	v1 = varargin{1}(:)';
+	v2 = varargin{2}(:);
+	orig_a = varargin{3};
+	
+	if (nargin == 4) shape = varargin{4}; endif
+    else
+	## usage: fftconv2(a, b[, shape])
+	
+	a = varargin{1};
+	b = varargin{2};
+	if (nargin == 3) shape = varargin{3}; endif
+
+    endif
+
+    if (rowcolumn)
+	a = fftconv2(orig_a, v2);
+	b = v1;
+    endif
+    
+    ra = rows(a);
+    ca = columns(a);
+    rb = rows(b);
+    cb = columns(b);
+
+    A = fft2(impad(a, [0 cb-1], [0 rb-1]));
+    B = fft2(impad(b, [0 ca-1], [0 ra-1]));
+
+    X = ifft2(A.*B);
+
+    if (rowcolumn)
+	rb = rows(v2);
+	ra = rows(orig_a);
+	cb = columns(v1);
+	ca = columns(orig_a);
+    endif
+    
+    if strcmp(shape,"same")
+	r_top = ceil((rb + 1) / 2);
+	c_top = ceil((cb + 1) / 2);
+	X = X(r_top:r_top + ra - 1, c_top:c_top + ca - 1);
+    elseif strcmp(shape, "valid")
+	X = X(rb:ra, cb:ca);
+    endif
+endfunction
+
+%!# usage: fftconv2(a,b,[, shape])
+%!shared a,b
+%! a = repmat(1:10, 5);
+%! b = repmat(10:-1:3, 7);
+%!assert(norm(fftconv2(a,b)-conv2(a,b)), 0, 1e6*eps)
+%!assert(norm(fftconv2(b,a)-conv2(b,a)), 0, 1e6*eps)
+%!assert(norm(fftconv2(a,b,'full')-conv2(a,b,'full')), 0, 1e6*eps)
+%!assert(norm(fftconv2(b,a,'full')-conv2(b,a,'full')), 0, 1e6*eps)
+%!assert(norm(fftconv2(a,b,'same')-conv2(a,b,'same')), 0, 1e6*eps)
+%!assert(norm(fftconv2(b,a,'same')-conv2(b,a,'same')), 0, 1e6*eps)
+%!assert(isempty(fftconv2(a,b,'valid')));
+%!assert(norm(fftconv2(b,a,'valid')-conv2(b,a,'valid')), 0, 1e6*eps)
+
+%!# usage: fftconv2(v1, v2, a[, shape])
+%!shared x,y,a
+%! x = 1:4; y = 4:-1:1; a = repmat(1:10, 5);
+%!assert(norm(fftconv2(x,y,a)-conv2(x,y,a)), 0, 1e6*eps)
+%!assert(norm(fftconv2(x,y,a,'full')-conv2(x,y,a,'full')), 0, 1e6*eps)
+%!assert(norm(fftconv2(x,y,a,'same')-conv2(x,y,a,'same')), 0, 1e6*eps)
+%!assert(norm(fftconv2(x,y,a,'valid')-conv2(x,y,a,'valid')), 0, 1e6*eps)
+
+%!demo
+%! ## Draw a cross
+%! N = 100;
+%! [x,y] = meshgrid(-N:N, -N:N);
+%! z = 0*x;
+%! z(N,1:2*N+1) = 1; z(1:2*N+1, N) = 1;
+%! imshow(z);
+%!
+%! ## Draw a sinc blob
+%! n = floor(N/10);
+%! [x,y] = meshgrid(-n:n, -n:n);
+%! b = x.^2 + y.^2; b = max(b(:)) - b; b = b / max(b(:));
+%! imshow(b);
+%!
+%! ## Convolve the cross with the blob
+%! imshow(real(fftconv2(z, b, 'same')*N))
+
+