彩色图像空间转换代码

图像篡改检测Matlab源代码说明（1）彩色图像空间转换代码（RGB空间转换为YUV空间）clc;clearall;closeall;img=imread('3.jpg');%figure;imshow(img);title('原图');img_5=img;img_ycbcr=rgb2ycbcr(img);%rgb-yuvfigure;subplot(121);imshow(img);title('原始图像');subplot(122);imshow(img_ycbcr);title('YUV空间图');图1彩色图像空间转换（2）图像分块（分为16*16子块）[row,col,i]=size(img_ycbcr);%对图像进行扩展row_expand=ceil(row/16)*16;%行数上取整再乘16，及扩展成16的倍数ifmod(row,16)~=0%行数不是16的倍数，用最后一行进行扩展fori=row:row_expandimg_ycbcr(i,:,:)=img_ycbcr(row,:,:);endendcol_expand=ceil(col/16)*16;%列数上取整ifmod(col,16)~=0%列数不是16的倍数，用最后一列进行扩展forj=col:col_expandimg_ycbcr(:,j,:)=img_ycbcr(:,col,:);endend（3）对各个分量（Y,Cr，Cb）进行离散余弦变换量化%对各分量进行4:2:0采样Y=img_ycbcr(:,:,1);%Y分量figure;subplot(231);imshow(Y);title('原Y分量');Cb=img_ycbcr(:,:,2);%Cb分量Cr=img_ycbcr(:,:,3);%Cr分量Cb=zeros(row_expand/2,col_expand/2);Cr=zeros(row_expand/2,col_expand/2);fori=1:row_expand/2forj=1:2:col_expand/2-1%奇数Cb(i,j)=double(img_ycbcr(i*2-1,j*2-1,2));Cr(i,j)=double(img_ycbcr(i*2-1,j*2+1,3));endendfori=1:row_expand/2%偶数forj=2:2:col_expand/2Cb(i,j)=double(img_ycbcr(i*2-1,j*2-2,2));Cr(i,j)=double(img_ycbcr(i*2-1,j*2,3));endendsubplot(232);imshow(uint8(Cb));title('原Cb分量');%Cb分量subplot(233);imshow(uint8(Cr));title('原Cr分量');%Cr分量Y_Table=[1611101624405161121214192658605514131624405769561417222951878062182237566810910377243555648110411392496478871031211201017292959811210010399];%亮度量化表CbCr_Table=[17,18,24,47,99,99,99,99;18,21,26,66,99,99,99,99;24,26,56,99,99,99,99,99;47,66,99,99,99,99,99,99;99,99,99,99,99,99,99,99;99,99,99,99,99,99,99,99;99,99,99,99,99,99,99,99;99,99,99,99,99,99,99,99];%色差量化表Qua_Factor=0.2;%量化因子%对Y分量进行处理Qua_Matrix=Qua_Factor*Y_Table;%Y量化矩阵Y=blkproc(Y,[88],'dct2(x)');%DCT变换Y=blkproc(Y,[88],'round(x./P1)',Qua_Matrix);%量化Y=blkproc(Y,[88],'x.*P1',Qua_Matrix);%反量化Y=blkproc(Y,[88],'idct2(x)');%反DCT变换subplot(234);imshow(uint8(Y));title('Y分量量化后');%对Cb分量进行处理Qua_Matrixcc=0.2*CbCr_Table;%Cb质量因子选取为5，并不采用Qua_Factor的数值Cb=blkproc(Cb,[88],'dct2(x)');%DCT变换Cb=blkproc(Cb,[88],'round(x./P1)',Qua_Matrixcc);%量化Cb=blkproc(Cb,[88],'x.*P1',Qua_Matrixcc);%反量化Cb=blkproc(Cb,[88],'idct2(x)');%反DCT变换subplot(235);imshow(uint8(Cb));title('Cb分量量化后');%对Cr分量进行处理Qua_Matrixcc=0.2*CbCr_Table;%Cr质量因子选取为5，并不采用Qua_Factor的数值Cr=blkproc(Cr,[88],'dct2(x)');%DCT变换Cr=blkproc(Cr,[88],'round(x./P1)',Qua_Matrixcc);%量化Cr=blkproc(Cr,[88],'x.*P1',Qua_Matrixcc);%反量化Cr=blkproc(Cr,[88],'idct2(x)');%反DCT变换subplot(236);imshow(uint8(Cr));title('Cr分量量化后');图2各个分量量化图像（4）对各分量（Y,Cr，Cb）进行二值化处理%对Y分量处理%差分处理dh_Y=diff(Y,[],2);%水平差分dvt_Y=diff(Y);%纵向差分dv_Y=dvt_Y';%显示功率谱密度fs=1000;n=0:1/fs:1;nfft=1024;window=blackman(100);noverlap=20;range='half';[pxx1,f]=pwelch(dh_Y,window,noverlap,nfft,fs,range);[pxx2,f]=pwelch(dvt_Y,window,noverlap,nfft,fs,range);plot_pxx1=10*log10(pxx1);plot_pxx2=10*log10(pxx2);figure,subplot(121);plot(f,plot_pxx1);title('水平功率谱密度');subplot(122);plot(f,plot_pxx2);title('垂直功率谱密度');%计算功率谱密度dh1_Y=blkproc(dh_Y,[88],'welch');%水平方向差分图像的区域功率谱密度dv1_Y=blkproc(dv_Y,[88],'welch');%垂直方向差分图像的区域功率谱密度b1_Y=blkproc(dh1_Y,[5131],'sum(x(:))');%水平方向差分图像的区域功率谱密度总和b2_Y=blkproc(dv1_Y,[5131],'sum(x(:))');%垂直方向差分图像的区域功率谱密度总和b_Y=b1_Y+b2_Y';%整幅图像的区域功率谱密度总和fori=1:size(b_Y,1)forj=1:size(b_Y,2)B_Y(i,j)=exp(log(sum((b_Y(i,j)+1)))/(sum(b_Y(i,j))+1);%整幅图像块效应评价endend%二值化处理T=exp(1.5)*sum(sum(B_Y))/((row_expand/8)*(col_expand/8));%阈值c_Y=zeros(size(Y));fori=1:size(Y,1)forj=1:size(Y,2)ifb_Y(ceil(i/8),ceil(j/8))Tc_Y(i,j)=1;endendendfigure;subplot(121);imshow(int8(b_Y));title('Y分量离散余弦系数图');subplot(122);imshow(c_Y);title('Y分量二值图像');图3Y分量离散余弦变换二值化图像图4Cb分量离散余弦变换二值化图像图5Cr分量离散余弦变换二值化图像(5)提取各个子图特征functionfeatureVec=GetBlockFeature(Block)featureVec=zeros(1,7);%特征数组，共设置7个特征R=Block(:,:,1);G=Block(:,:,2);B=Block(:,:,3);%%cl,c2,c3三个特征值featureVec(1)=mean2(R(:));featureVec(2)=mean2(G(:));featureVec(3)=mean2(B(:));%%c4..5..6..7四个特征值,Y通道公式Y=im2double(0.299*R+0.587*G+0.114*B);%块内所有像素灰度之和totalGray=sum(Y(:));%c4sum_partI=sum(sum(Y(1:8,:)));featureVec(4)=sum_partI/totalGray;%c5sum_partI=sum(sum(Y(:,1:8)));featureVec(5)=sum_partI/totalGray;%c6sumPart11=0.0;forr=1:16forc=r:16sumPart1=sumPart11+Y(r,c);endendfeatureVec(6)=sumPart1/totalGray;%c7sumPart11=0.0;forr=1:16forc=1:(16-r+1)sumPart1=sumPart11+Y(r,c);endendfeatureVec(7)=sumPart1/totalGray;return;(6)特征排序FeatureVecThresh=[2.5,1.5,3.0,0.006,0.005,0.005,0.005];BlockDistThresh=10;BlockPairsNum=NumBlocks-1;ShiftVectors=zeros(1,2);candidateSimilarPairs.indices=zeros(1,2);candidateSimilarPairs.frequence(1)=0;[features,indices]=sortrows(Blocks.features);vi=0;fori=1:NumBlocks-1ind1=indices(i);ind2=indices(i+1);deltaCoord=Blocks.topLeft(ind1,:)-Blocks.topLeft(ind2,:);deltaFeature=Blocks.features(ind1,:)-Blocks.features(ind2,:);pixelDist=norm(deltaCoord);%向景取模if(pixelDistBlockDistThresh)&&(sum(abs(deltaFeature)FeatureVecThresh)==7)vi=vi+1;candidateSimilarPairs.indices(vi,:)=[ind1ind2];ShiftVectors(vi,:)=Blocks.topLeft(ind1,:)-Blocks.topLeft(ind2,:);endend(7)查找并删除重复图像块SimilarPairs.indices=zeros(1,2);pairs.vector{1}=ShiftVectors(1,:);%第一行pairs.freq(1)=1;ind=1;fori=2:size(ShiftVectors,1);vec=ShiftVectors(i,:);isExist=fals