function [X,map] = im2ind(R,G,B)
%IM2IND Convert image to indexed image with unique colors in map.
%
% IM2IND converts an RGB, intensity or indexed image to an indexed
% image with a color map where duplicates and unused colors are
% removed. No color quantization is done as with CMUNIQUE in the image
% processing toolbox. The number of rows in the returned color map is
% the number of unique colors in the image.
%
% [X,MAP] = IM2IND(R,G,B) converts the RGB image in the matrices R, G
% and B to an indexed image X with color map MAP.
%
% [X,MAP] = IM2IND(I) converts the intensity image I to an indexed
% image X with color map MAP.
%
% [Y,NEWMAP] = IM2IND(X,MAP) converts the indexed image X with color
% map MAP to an indexed image Y with color map NEWMAP.
%
% The conversion is done by sorting and comparing consecutive
% elements. IM2IND is much faster than IM2INDS but uses more memory.
%
% See also IM2INDS, IM2INDD.
% Author: Peter J. Acklam
% Time-stamp: 1998-04-25 01:16:55
% E-mail: jacklam@math.uio.no (Internet)
% URL: http://www.math.uio.no/~jacklam
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
% Check input arguments.
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
error( nargchk( 1, 3, nargin ) );
if nargin == 2 % IM2IND(X,MAP)
if size( G, 2 ) ~= 3
error( 'Color map must have three columns.' );
end
elseif nargin == 3 % IM2IND(R,G,B)
if any( size(R) ~= size(G) ) | any( size(G) ~= size(B) )
error( 'R, G, and B must have the same size' );
end
end
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
% The code below uses some recycling of variables to save memory.
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
if nargin == 1 % IM2IND(I)
[ rows, cols ] = size( R ); % Size of image.
pixels = rows * cols; % Number of pixels in image.
[ R, idx ] = sort( R(:) ); % Sort intensity values.
% This vector contains ones where the unique intensity values are.
isunique = logical( ones( pixels, 1 ) );
isunique(2:pixels) = R(1:pixels-1) ~= R(2:pixels);
% Build color map.
map = R(isunique); % Color map with unique intensity
map = map(:,ones(1,3)); % values.
clear R; % Free some memory.
% Build index matrix.
X = zeros( rows, cols ); % Initialize output index matrix.
X(idx) = cumsum( isunique ); % Fill indices into matrix.
elseif nargin == 2 % IM2IND(X,MAP)
[ rows, cols ] = size( R ); % Size of image.
pixels = rows * cols; % Number of pixels in image.
colors = size( G, 1 ); % Number of colors in map.
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
% See if the indexed image is valid.
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
if max( R(:) ) > colors
error( 'Maximum index value exceeds number of colors in map.' );
end
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
% Rename variables. It might have saved some memory using R and G
% directly rather than X and map, but it would be more confusing.
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
X = R; clear R;
map = G; clear G;
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
% Sort the color map and build index vectors.
% idx: original color map -> sorted color map
% idxinv: sorted color map -> original color map
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
% Color of pixel (i,j): map( X(i,j) , : )
% Find index vector that maps from original to sorted color map.
[ dummy, ind ] = sort( map( : , 3 ) ); idx = ind;
[ dummy, ind ] = sort( map( idx, 2 ) ); idx = idx(ind);
[ dummy, ind ] = sort( map( idx, 1 ) ); idx = idx(ind);
% Sort color map and find the "inverse" index vector.
map = map(idx,:);
idxinv = zeros(size(idx));
idxinv(idx) = 1:colors;
% Color of pixel (i,j): map( idxinv( X(i,j) ) , : )
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
% Remove duplicate colors from color map.
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
% This vector contains ones where the unique colors are.
isunique = logical( ones( colors, 1 ) );
isunique(2:colors) = map(1:colors-1,1) ~= map(2:colors,1) ...
| map(1:colors-1,2) ~= map(2:colors,2) ...
| map(1:colors-1,3) ~= map(2:colors,3);
map = map(isunique,:); % Remove duplicates.
% This "index reduction vector" creates a new index vector with the
% smalles possible values, e.g., [ 2 2 5 8 8 8 ] -> [ 1 1 2 3 3 3 ].
idxrdc = ( 1:colors )' - cumsum( ~isunique );
X(:) = idxrdc(idxinv(X)); % Fill in the indices.
% Color of pixel (i,j): map( X(i,j) , : )
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
% Remove unused colors from color map. CMUNIQUE actually calls
% IMHIST to help with this job, and IMHIST calls a MEX program!
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
colors = size( map, 1 ); % Number of colors in new map.
v = logical( zeros( colors, 1 ) ); % Initialize to zero.
v(X) = 1; % Mark used colors.
map = map(v,:); % Remove unused colors.
v = cumsum(v); % Make an index mapping vector.
X = v(X); % New index matrix.
else % IM2IND(R,G,B)
[ rows, cols ] = size( R ); % Size of image.
pixels = rows * cols; % Number of pixels in image.
G = [ R(:) G(:) B(:) ]; % Create a giant color map.
clear R B; % Free some memory.
% Sort the rows in the color map.
[ dummy, ind ] = sort( G( : , 3 ) );
idx = ind;
[ dummy, ind ] = sort( G( idx, 2 ) );
idx = idx(ind);
[ dummy, ind ] = sort( G( idx, 1 ) );
idx = idx(ind);
G = G(idx,:);
clear dummy ind; % Free some memory.
% This vector contains ones where the unique colors are.
isunique = logical( ones( pixels, 1 ) );
isunique(2:pixels) = G(1:pixels-1,1) ~= G(2:pixels,1) ...
| G(1:pixels-1,2) ~= G(2:pixels,2) ...
| G(1:pixels-1,3) ~= G(2:pixels,3);
% Build color map.
map = G(isunique,:); % Color map with unique colors.
clear G; % Free some memory.
% Build index matrix.
X = zeros( rows, cols ); % Initialize output index matrix.
X(idx) = cumsum( isunique ); % Fill indices into matrix.
end
