random41.rar_random

%Blind embedding and linear correlation detection clear all; %Generate a reference pattern the same size as the images wr=randn(112,92); tm=mean2(wr); wr=wr-tm;%zero mean ts=std2(wr); wr=wr/ts;%unit variance %Show the reference pattern figure imshow(wr,[]); %400 face images, 112 x 92 pixels, in 40 folders of 10 images each numclass=40; numface=10; for s=1:numclass for i=1:numface %Read image c0=double(imread(strcat('C:\facedb_bmp\s',num2str(s),'\',num2str(i),'.bmp'))); %Add reference pattern to the image to embed a 1 wa=double(c0+wr); %Add negative reference pattern to the image to embed a 0 nwr=wr*-1; nwa=double(c0+nwr); %Calculate linear correlation for images carrying a 1, a 0, %or no watermark and store it into a vector corr1(10*(s-1)+i)=sum(sum(wa.*wr))/(112*92); corr0(10*(s-1)+i)=sum(sum(nwa.*wr))/(112*92); corrN(10*(s-1)+i)=sum(sum(c0.*wr))/(112*92); end end %Calculate the histograms for the detection-value vectors [a1,b1]=hist(corr1,-3:0.1:3); [a2,b2]=hist(corr0,-3:0.1:3); [a3,b3]=hist(corrN,-3:0.1:3); %Plot the histograms figure plot(b2,a2/400*100,'red'); gtext('m=0'); hold; plot(b3,a3/400*100,'green'); gtext('No watermark'); plot(b1,a1/400*100,'blue'); gtext('m=1'); xlabel('Detection value'); ylabel('Percentage of images');