Signed distance function vectorization

[fotios]

Hello list!

So here is my problem (weakly coupled to Octave). Given a 
non-self-intersecting closed curve (simply connected) and a grid 
comprised by square elements, I would like to get the signed distance 
function from the centers of those elements on that grid. Currently the 
way I do it is shown below

function [ d ] = sdistf ( cx, cy, x, y )

n = [ diff( y ), -diff( x ) ];
n = (n ./ repmat( sqrt( n(:, 1).^2 + n(:, 2).^2 ), 1, 2 ));
r = [ x(1:(end - 1)) - cx, y(1:(end - 1)) - cy ]';
d = min( diag( n*r ) );
endfucntion

mytest
# yada yada
for row = 1 : numel( cx )
  for col = 1 : numel (cx)
    d(row, col) = sdistf ( xc(row, col), yc(row, col), px, py );
  end
end
# bla bla

and I was able to vectorize everything in the worst (memory-wise) way . 
. . a clever way of doing this in Octave (without using level-sets or 
anything like that, just an vectorized version of the above). As a 
result of my try to find some info - or a starting point for you that 
will try to help me - I found that one

http://statinfer.wordpress.com/2011/11/14/efficient-matlab-i-pairwise-distances/

/Fotis