laplacian score for blocks of image

I am working on feature extraction of faces image. I have to select the most relevant features by assigning a score for each feature then rearrange them according to their importance. I am trying to adapt the code in order to give a laplacian score for each block of image. and then rearrage these blocks according to their importance and select the most relevant blocks. So how can I adapt this code?!

here is the code:

% ====================================================== % % Feature selection based on supervised Laplacian scores % % Framework numero 1 % ====================================================== %

x=TrainDBN; y=TestDBN; % % %%==========parametrage==========%% AvecPCA=1; % Si 1 alors Appliquer PCA sur la base d'apprentissage PCA_Ratio=95; ClassifClassique=1; % Si 1 faire la classification classique % % qui prend en compte tous les dimensions % PourcentagePixels=20;

K_KNN=1; K_heterogenes=3; K_homogenes=5;

%%============PCA(Princomp)===============%%

if AvecPCA==1 [x,y]=mypca(x,y,PCA_Ratio); end

% Calculer le Kernel Heat disp('Computing the distances ...') [N M]=size(x.mat); % DXiXj est une matrice contenant les distances entre XI et Xj DXiXj=zeros(N);

SommedXi_Xj=0; for i=1:N for j=1:N dXi_Xj=(norm(x.mat(i,:)-x.mat(j,:)))^2; DXiXj(i,j)=dXi_Xj; SommedXi_Xj=SommedXi_Xj+dXi_Xj; end end

% sigma determines the bandwidth of the Gaussian kernel. sigma2=SommedXi_Xj/N^2; T=2*sigma2;

% Compute the Laplacian matrix Lb (heterogenous samples) disp('Compute the Laplacian matrix Lb ...') [Wb Db Lb] = LbMaker(x.mat', x.Classes', K_heterogenes, sigma2);

% Compute the Laplacian matrix Lw (homogenous samples) disp('Compute the Laplacian matrix Lw ...') [Ww Dw Lw] = LwMaker(x.mat', x.Classes', K_homogenes, sigma2);

% Compute the importance score L disp('Compute the importance score L ...') L=[]; for r=1:M f_r=x.mat(:,r); L=[L;(f_r'*Lb*f_r)/(f_r'*Lw*f_r)]; end

[L_sorted idx]=sort(L,'descend'); x.mat=x.mat(:,idx); y.mat=y.mat(:,idx);

if ClassifClassique ==1 To=myclassify(x,y,K_KNN); else To=myclassify1(x,y,K_KNN,PourcentagePixels); end % [maxi indx]=max(To(:,2)); % [To(indx,1) maxi]

% clear x y;