Gray-level run-length.zip_Gray-level run_run_游程矩阵_灰度游程矩阵_灰度游程矩阵特

function [glrlm_45, feature_45]= degree_45(SI) [r,c]=size(SI); SI=rot90(SI); % ratate the matrix of 90 degree a=-1*ones(c-1,r+1); % create a zero c-1 by r+1 matrix for i=1:c-1 a(c-i,1:length((diag(SI,i))'))= (diag(SI,i))'; end % extract the element in the main diagonal d=(diag(SI,0))'; % extract the diagonal elements below the main diagonal a1=-1*ones(c-1,r+1); for i=1:c-1 a1(i,1:length((diag(SI',i))'))=(diag(SI',i))'; end SI=[a;[d,-1];a1]; graylevel=16; glrlm_45 = zeros(graylevel,r); % row represents the number of gray levels, column represents the number of differnt run lengths that occur for i = 1:2*r-1 % the total numberow of row becomes r+(r-1) y=SI(i,:); % extract each row of sacled image % run length Encode of each vector index = [ find(y(1:end-1) ~= y(2:end)), length(y) ]; len = diff([ 0 index ]); % run lengths val = y(index); % run values temp = accumarray([val(1:end-1);len(1:end-1)]',1,[graylevel r]);% compute current numbers (or contribution) for each bin in GLRLM glrlm_45 = temp + glrlm_45; % accumulate each contribution end P=sum(sum(glrlm_45)); % total number of runs SRE=0; LRE=0; LGRE=0; HGRE=0;SRLGE=0; SRHGE=0; LRLGE=0;LRHGE=0; for i = 1:graylevel for j = 1:c SRE=SRE+glrlm_45(i,j)/j^2 ; % short run emphasis LRE=LRE+glrlm_45(i,j)*j^2 ; % long run emphasis LGRE=LGRE+glrlm_45(i,j)/i^2; HGRE=HGRE+glrlm_45(i,j)*i^2; SRLGE=SRLGE+glrlm_45(i,j)/(i^2*j^2); SRHGE=SRHGE+glrlm_45(i,j)*i^2/j^2; LRLGE=LRLGE+glrlm_45(i,j)*j^2/i^2; LRHGE=LRHGE+glrlm_45(i,j)*j^2*i^2; end end SRE=SRE/P; LRE=LRE/P; LGRE=LGRE/P; HGRE=HGRE/P; SRLGE=SRLGE/P; SRHGE=SRHGE/P; LRLGE= LRLGE/P;LRHGE=LRHGE/P; % Gray-level Nonuniformity GLN=0; for j=1:c GLN=GLN+sum(glrlm_45(:,j))^2; end GLN=GLN/P; % Run length nonuniformity RLN=0; for i=1:graylevel RLN=RLN+sum(glrlm_45(i,:))^2; end RLN=RLN/P; RP = P / (graylevel*c); % run percentage feature_45=[SRE,LRE,GLN,RLN,RP,LGRE,HGRE,SRLGE,SRHGE,LRLGE,LRHGE];