function ROC ( X, GoodToBad, xAxis, lineattr )

%
%  function ROC ( X, GoodToBad, xAxis ) - plots a ROC curve.
%
%
%
%  Input data:
%
%  	X	N x 2	This is the main input variable. The first column
%               	contains the score of each image, the second column contains the
%			labels. Noise (images that don't contain the
% 			object) is labeled with 0, signal (images that
%                       contain the object) is labeled with k > 0. Noise
%                       is supposed to have a low score, signal is
%                       supposed to have a high score. N is the number of
%                       samples.
%
%%%%%%
%
%  GoodToBad	scalar 	Allows to rescale the ratio of signal and noise samples. Just 
% 			put 1 here if you're not sure.
%
%  xAxis	2 x 1	The range for the x-axis. This allows to close in or zoom out. 
%			Put [0 1] if you are not sure.
%
%  lineattr	string	These are the attributes that will be used for the ROC curve. Use
%			this if you want to plot several curves in the same plot.
%		
%
%  (c) 1997 by California Institute of Technology
% 
%  Markus Weber
%
%  edited P.Moreels, 2004.

if nargin < 4,

	lineattr = 'r-';
end

if nargin == 1
   
   GoodToBad = 1;
   xAxis = [0 1];
   
end


N_SAMPLES = size(X, 1);
N_SIGNAL  = length(find(X(:, 2)));
N_NOISE   = N_SAMPLES - N_SIGNAL;

[D, I] = sort(rand(N_SAMPLES, 1));

X = X(I, :);

[D, I] = sort(X(:, 1));

X = flipud(X(I, :));

P = cumsum([ (X(:, 2) > 0) / N_SIGNAL, (X(:, 2) == 0) / N_NOISE ]);

P = [0 0;P];

plot(P(:, 2) * 1/GoodToBad, P(:, 1), lineattr); grid;

set(gca, 'xtick', xAxis(1):(xAxis(2) - xAxis(1)) / 5:xAxis(2), 'ytick', 0:0.1:1);

axis ([xAxis(:)', 0 1 ]);

xlabel('false alarm rate');
ylabel('detection rate');
title('ROC curve');