基于DCT变换的图像水印方法

书本中提供的方法，我总是无法得到很好的效果。也许是我的小改出现了问题吧。

            carrierI=imread('cameraman.tif');
          
            watermarkI=imread('21.bmp');
          
            watermarkI=imbinarize(watermarkI);
          
            %定义DCT中频系数
          
            DCT_coef=[
          
             0,0,0,1,1,1,1,0;
          
             0,0,1,1,1,1,0,0;
          
             0,1,1,1,1,0,0,0;
          
             1,1,1,1,0,0,0,0;
          
             1,1,1,0,0,0,0,0;
          
             1,1,0,0,0,0,0,0;
          
             1,0,0,0,0,0,0,0;
          
             0,0,0,0,0,0,0,0
          
            ];
          
            %设置水印强度
          
            k=20;

嵌入水印

%设置相同的随机数生成器状态J,作为检测时的系统密钥K

rand('state',7);

pn_sequence_zero = round(rand( 1, sum(DCT_coef(:))));

watermarkedI=DCT_encode(carrierI,watermarkI,DCT_coef,k,pn_sequence_zero);

%显示峰值信噪比

xsz =255 * 255 * numel(carrierI)/ sum( sum( ( carrierI - watermarkedI).^2));

psnr= 10 * log10( xsz);

fprintf("图像失真程度指标：峰值信噪比PSNR=%g",psnr);

                图像失真程度指标：峰值信噪比PSNR=48.5722
              

figure;

subplot(1,2,1);

imshow(carrierI);

title("载体图像");

subplot(1,2,2);

imshow(watermarkedI);

title("带有水印的图像")

提取水印

%设置相同的随机数生成器状态J,作为检测时的系统密钥K

rand('state',7);

pn_sequence_zero = round(rand( 1, sum(DCT_coef(:))));

watermark2I=DCT_decode(watermarkedI,carrierI,DCT_coef,pn_sequence_zero);

%计算提取的水印和原始水印的相似程度

sim = corr2( watermarkI,watermark2I );

fprintf("相似度%g",sim);

                相似度-0.756504
              

figure();

subplot(1,2,1);

imshow(watermarkI);

title('原始水印')

subplot(1,2,2)

imshow(watermark2I==0);

title('还原出来的水印')