glcm_features2.m

segmentation code in image processing

function [out] = GLCM_Features1(glcmin,pairs)
% 
% GLCM_Features2 helps to calculate the features from the different GLCMs
% that are input to the function. The GLCMs are stored in a i x j x n
% matrix, where n is the number of GLCMs calculated usually due to the
% different orientation and displacements used in the algorithm. Usually
% the values i and j are equal to 'NumLevels' parameter of the GLCM
% computing function graycomatrix(). Note that matlab quantization values
% belong to the set {1,..., NumLevels} and not from {0,...,(NumLevels-1)}
% as provided in some references
% http://www.mathworks.com/access/helpdesk/help/toolbox/images/graycomatrix
% .html
% 
% This vectorized version of GLCM_FEatures1.m reduces the 19 'for' loops
% used in the earlier code to 5 'for' loops
% http://blogs.mathworks.com/loren/2006/07/12/what-are-you-really-measuring
% /
% Using tic toc and cputime as in above discussion I find no significant
% improvement even after reducing all the loops. If anyone can figure out
% why, please let me know. Thanks!
%
% Although there is a function graycoprops() in Matlab Image Processing
% Toolbox that computes four parameters Contrast, Correlation, Energy,
% and Homogeneity. The paper by Haralick suggests a few more parameters
% that are also computed here. The code is not fully vectorized and hence
% is not an efficient implementation but it is easy to add new features
% based on the GLCM using this code. Takes care of 3 dimensional glcms
% (multiple glcms in a single 3D array)
% 
% If you find that the values obtained are different from what you expect 
% or if you think there is a different formula that needs to be used 
% from the ones used in this code please let me know. 
% A few questions which I have are listed in the link 
% http://www.mathworks.com/matlabcentral/newsreader/view_thread/239608
%
%
%
% Features computed 
% Autocorrelation: [2]                      (out.autoc)
% Contrast: matlab/[1,2]                    (out.contr)
% Correlation: matlab                       (out.corrm)
% Correlation: [1,2]                        (out.corrp)
% Cluster Prominence: [2]                   (out.cprom)
% Cluster Shade: [2]                        (out.cshad)
% Dissimilarity: [2]                        (out.dissi)
% Energy: matlab / [1,2]                    (out.energ)
% Entropy: [2]                              (out.entro)
% Homogeneity: matlab                       (out.homom)
% Homogeneity: [2]                          (out.homop)
% Maximum probability: [2]                  (out.maxpr)
% Sum of squares: Variance [1]              (out.sosvh)
% Sum average [1]                           (out.savgh)
% Sum variance [1]                          (out.svarh)
% Sum entropy [1]                           (out.senth)
% Difference variance [1]                   (out.dvarh)
% Difference entropy [1]                    (out.denth)
% Information measure of correlation1 [1]   (out.inf1h)
% Informaiton measure of correlation2 [1]   (out.inf2h)
% Inverse difference (INV) is homom [3]     (out.homom)
% Inverse difference normalized (INN) [3]   (out.indnc) 
% Inverse difference moment normalized [3]  (out.idmnc)
%
% The maximal correlation coefficient was not calculated due to
% computational instability 
% http://murphylab.web.cmu.edu/publications/boland/boland_node26.html
%
% Formulae from MATLAB site (some look different from
% the paper by Haralick but are equivalent and give same results)
% Example formulae: 
% Contrast = sum_i(sum_j(  (i-j)^2 * p(i,j) ) ) (same in matlab/paper)
% Correlation = sum_i( sum_j( (i - u_i)(j - u_j)p(i,j)/(s_i.s_j) ) ) (m)
% Correlation = sum_i( sum_j( ((ij)p(i,j) - u_x.u_y) / (s_x.s_y) ) ) (p[2])
% Energy = sum_i( sum_j( p(i,j)^2 ) )           (same in matlab/paper)
% Homogeneity = sum_i( sum_j( p(i,j) / (1 + |i-j|) ) ) (as in matlab)
% Homogeneity = sum_i( sum_j( p(i,j) / (1 + (i-j)^2) ) ) (as in paper)
% 
% Where:
% u_i = u_x = sum_i( sum_j( i.p(i,j) ) ) (in paper [2])
% u_j = u_y = sum_i( sum_j( j.p(i,j) ) ) (in paper [2])
% s_i = s_x = sum_i( sum_j( (i - u_x)^2.p(i,j) ) ) (in paper [2])
% s_j = s_y = sum_i( sum_j( (j - u_y)^2.p(i,j) ) ) (in paper [2])
%
% 
% Normalize the glcm:
% Compute the sum of all the values in each glcm in the array and divide 
% each element by it sum
%
% Haralick uses 'Symmetric' = true in computing the glcm
% There is no Symmetric flag in the Matlab version I use hence
% I add the diagonally opposite pairs to obtain the Haralick glcm
% Here it is assumed that the diagonally opposite orientations are paired
% one after the other in the matrix
% If the above assumption is true with respect to the input glcm then
% setting the flag 'pairs' to 1 will compute the final glcms that would result 
% by setting 'Symmetric' to true. If your glcm is computed using the
% Matlab version with 'Symmetric' flag you can set the flag 'pairs' to 0
%
% References:
% 1. R. M. Haralick, K. Shanmugam, and I. Dinstein, Textural Features of
% Image Classification, IEEE Transactions on Systems, Man and Cybernetics,
% vol. SMC-3, no. 6, Nov. 1973
% 2. L. Soh and C. Tsatsoulis, Texture Analysis of SAR Sea Ice Imagery
% Using Gray Level Co-Occurrence Matrices, IEEE Transactions on Geoscience
% and Remote Sensing, vol. 37, no. 2, March 1999.
% 3. D A. Clausi, An analysis of co-occurrence texture statistics as a
% function of grey level quantization, Can. J. Remote Sensing, vol. 28, no.
% 1, pp. 45-62, 2002
% 4. http://murphylab.web.cmu.edu/publications/boland/boland_node26.html
%
%
% Example:
%
% Usage is similar to graycoprops() but needs extra parameter 'pairs' apart
% from the GLCM as input
% I = imread('circuit.tif');
% GLCM2 = graycomatrix(I,'Offset',[2 0;0 2]);
% stats = GLCM_features2(GLCM2,0)
% The output is a structure containing all the parameters for the different
% GLCMs
%
% [Avinash Uppuluri: [e-mail]: Last modified: 12/07/08]

% If 'pairs' not entered: set pairs to 0 
if ((nargin > 2) || (nargin == 0))
   error('Too many or too few input arguments. Enter GLCM and pairs.');
elseif ( (nargin == 2) ) 
    if ((size(glcmin,1) <= 1) || (size(glcmin,2) <= 1))
       error('The GLCM should be a 2-D or 3-D matrix.');
    elseif ( size(glcmin,1) ~= size(glcmin,2) )
        error('Each GLCM should be square with NumLevels rows and NumLevels cols');
    end    
elseif (nargin == 1) % only GLCM is entered
    pairs = 0; % default is numbers and input 1 for percentage
    if ((size(glcmin,1) <= 1) || (size(glcmin,2) <= 1))
       error('The GLCM should be a 2-D or 3-D matrix.');
    elseif ( size(glcmin,1) ~= size(glcmin,2) )
       error('Each GLCM should be square with NumLevels rows and NumLevels cols');
    end    
end


format long e
if (pairs == 1)
    newn = 1;
    for nglcm = 1:2:size(glcmin,3)
        glcm(:,:,newn)  = glcmin(:,:,nglcm) + glcmin(:,:,nglcm+1);
        newn = newn + 1;
    end
elseif (pairs == 0)
    glcm = glcmin;
end

size_glcm_1 = size(glcm,1);
size_glcm_2 = size(glcm,2);
size_glcm_3 = size(glcm,3);

% checked 
out.autoc = zeros(1,size_glcm_3); % Autocorrelation: [2] 
out.contr = zeros(1,size_glcm_3); % Contrast: matlab/[1,2]
out.corrm = zeros(1,size_glcm_3); % Correlation: matlab
out.corrp = zeros(1,size_glcm_3); % Correlation: [1,2]
out.cprom = zeros(1,size_glcm_3); % Cluster Prominence: [2]
out.cshad = zeros(1,size_glcm_3); % Cluster Shade: [2]
out.dissi = zeros(1,size_glcm_3); % Dissimilarity: [2]
out.energ = zeros(1,size_glcm_3); % Energy: matlab / [1,2]
out.entro = zeros(1,size_glcm_3); % Entropy: [2]
out.homom = zeros(1,size_glcm_3); % Homogeneity: matlab
out.homop = zeros(1,size_glcm_3); % Homogeneity: [2]
out.maxpr = zeros(1,size_glcm_3); % Maximum probability: [2]

out.sosvh = zeros(1,size_glcm_3); % Sum of sqaures: Variance [1]
out.savgh = zeros(1,size_glcm_3); % Sum average [1]
out.svarh = zeros(1,size_glcm_3); % Sum variance [1]
out.senth = zeros(1,size_glcm_3); % Sum entropy [1]
out.dvarh = zeros(1,size_glcm_3); % Difference variance [4]
%out.dvarh2 = zeros(1,size_glcm_3); % Difference variance [1]
out.denth = zeros(1,size_glcm_3); % Difference entropy [1]
out.inf1h = zeros(1,size_glcm_3); % Information measure of correlation1 [1]
out.inf2h = zeros(1,size_glcm_3); % Informaiton measure of correlation2 [1]
%out.mxcch = zeros(1,size_glcm_3);% maximal correlation coefficient [1]
%out.invdc = zeros(1,size_glcm_3);% Inverse difference (INV) is homom [3]
out.indnc = zeros(1,size_glcm_3); % Inverse difference normalized (INN) [3]
out.idmnc = zeros(1,size_glcm_3); % Inverse difference moment normalized [3]

glcm_sum  = zeros(size_glcm_3,1);