%LABELIM Construct image of object (pixel) labels
%
% IM = LABELIM(A)
% IM = A*LABELIM
%
% INPUT
% A Dataset containing images stored as features
%
% OUTPUT
% IM Image containing the labels of the objects
%
% DESCRIPTION
% For a dataset A containing images stored as features, where each pixel
% corresponds to an object, the routine creates an image presenting the
% labels of the objects. Note that if the number of classes is small, e.g.
% 2, an appropriate colormap will have to be loaded for displaying the
% result using IMAGE(LABELS). More appropriate, LABELS should be multiplied
% such that the minimum and maximum of LABELS are well spread in the [1,64]
% interval of the standard colormaps. bb
% The resulting image may also directly be displayed by:
% LABELIM(A) or
% A*LABELIM
% for which a suitable colormap is loaded automatically.
%
% SEE ALSO (<a href="http://37steps.com/prtools">PRTools Guide</a>)
% DATASETS, CLASSIM
% Copyright: R.P.W. Duin, duin@ph.tn.tudelft.nl
% Faculty of Applied Sciences, Delft University of Technology
% P.O. Box 5046, 2600 GA Delft, The Netherlands
% $Id: labelim.m,v 1.2 2006/03/08 22:06:58 duin Exp $
function labels = labelim(a)
% No arguments given: return an untrained mapping.
if (nargin == 0)
labels = prmapping('labelim','fixed');
return
end
isfeatim(a); % Assert that A is a feature image dataset.
[n,m] = getobjsize(a); % Get image size and reshape labels to image.
J = getnlab(a); labels = reshape(J,n,m);
if (nargout == 0)
n = 61/(size(a,2)+0.5); % If no output is requested, display the
imagesc(labels*n); % image with a suitably scaled colormap.
colormap colorcube;
clear labels;
end
return