I am getting an error in watermark extraction.. actually extracted watermark is showing blank image. please help me with this code.. thank you

% Step 1: Read the original color image OI of size N x N
OI = imread('mandrill.jpg');
OI = uint8(imresize(OI, [512 512]));

% Step 2: Select the Red color component from the original image
color_component = OI(:, :, 1); % Red color component

% Step 3: Apply DWT to decompose the color component into LL, HL, LH, and HH sub-bands
[LL, HL, LH, HH] = dwt2(color_component, 'haar');

% Step 4: Apply DWT to the LL sub-band to decompose it further
[LL_LL, LL_HL, LL_LH, LL_HH] = dwt2(LL, 'haar');

% Step 5: Apply DCT to the LL_HH sub-band and obtain the DCT coefficient matrix B
B = dct2(LL_HH);

% Step 6: Apply SVD to B and obtain U, S, and V
[U, S, V] = svd(B);

% Step 7: Read the watermark image OW of size N/2 x N/2
OW = imread('peppers.jpg');
OW = uint8(imresize(OW, [256 256]));
OW = OW(:, :, 1); % Red color component
% Step 8: Apply DWT to decompose the watermark image into WLL, WHL, WLH, and WHH sub-bands
[WLL, WHL, WLH, WHH] = dwt2(OW, 'haar');

% Step 9: Apply DCT to the WHH sub-band and obtain the DCT coefficient matrix D
D = dct2(WHH);

% Step 10: Apply SVD to D and obtain U1, S1, and V1
[U1, S1, V1] = svd(D);

% Step 11: Modify S with the watermark by calculating S2 = S + alpha * S1
alpha = 0.01; % Modify the value of alpha as needed
S2 = S + alpha * S1;

% Step 12: Obtain B* using B* = U * S2 * V'
B_modified = U * S2 * V';

% Step 13: Apply inverse DCT to B* to produce the modified LL_HH* sub-band
LL_HH_modified = idct2(B_modified);

% Step 14: Apply inverse DWT to LL_LL, LL_HL, LL_LH, and LL_HH* to get the modified LL* sub-band
LL_modified = idwt2(LL_LL, LL_HL, LL_LH, LL_HH_modified, 'haar');

% Step 15: Apply inverse DWT to LL*, HL, LH, and HH to get the watermarked image
WI = idwt2(LL_modified, HL, LH, HH, 'haar');

% Step 16: Set the modified color component to the original image
OI_watermarked = OI;
OI_watermarked(:, :, 1) = WI; % Replace the Red component with the watermarked component

% Display and save the watermarked image
figure;
imshow(OI_watermarked);
title('Watermarked Image');
imwrite(OI_watermarked, 'watermarked_image.jpg');
% Step 1: Read the watermarked image
OI_watermarked = imread('watermarked_image.jpg');

% Step 2: Select the Red color component from the watermarked image
color_component_watermarked = OI_watermarked(:, :, 1); % Red color component

% Step 3: Apply DWT to decompose the color component into LL, HL, LH, and HH sub-bands
[LL_watermarked, HL_watermarked, LH_watermarked, HH_watermarked] = dwt2(color_component_watermarked, 'haar');

% Step 4: Apply DWT to the LL sub-band to decompose it further
[LL_LL_watermarked, LL_HL_watermarked, LL_LH_watermarked, LL_HH_watermarked] = dwt2(LL_watermarked, 'haar');

% Step 5: Select the LL_HH* band and apply DCT to it to obtain the DCT coefficient matrix A
A = dct2(LL_HH_watermarked);

% Step 6: Apply SVD to A and obtain WU, WS, and WV
[WU, WS, WV] = svd(A);

% Step 7: Calculate Sr by subtracting the original singular values WS from S
Sr = (S - WS) / alpha;

% Step 8: Obtain Wr using Wr = U1 * Sr * V1'
Wr = U1 * Sr * V1';

% Step 9: Apply inverse DCT to Wr to produce the extracted LL_HH* sub-band
LL_HH_extracted = idct2(Wr);

% Step 10: Apply inverse DWT to LL_LL_watermarked, LL_HL_watermarked, LL_LH_watermarked, and LL_HH_extracted to get the extracted LL* sub-band
LL_extracted = idwt2(LL_LL_watermarked, LL_HL_watermarked, LL_LH_watermarked, LL_HH_extracted, 'haar');

% Step 11: Apply inverse DWT to LL_extracted, HL_watermarked, LH_watermarked, and HH_watermarked to get the extracted watermarked image
OW_extracted = idwt2(LL_extracted, HL_watermarked, LH_watermarked, HH_watermarked, 'haar');

% Set the extracted watermark to the original watermark size
extracted_watermark = imresize(OW_extracted, [256 256]);

% Display the extracted watermark image
figure;
imshow(extracted_watermark);
title('Extracted Watermark');

% Save the extracted watermark image
imwrite(extracted_watermark, 'extracted_watermark.jpg');

Answer 1

[Image Analyst]

Looks like it requires teh Wavelet toolbox, which I don't have. Is extracted_watermark floating point? If so, try this:

imshow(extracted_watermark, []);

[Suniti Singh]

sir now in the extraction part i am able to extract the image but the extracted image is not watermark image it comes as input image i.e mandrill. also PSNR is showing infinte between input image and watermarked image. can you please help me out with this. how i can resolve this.

Thank you so much for your reply.

[Image Analyst]

Here's some help:

Debugging in MATLAB | Doug's MATLAB Video Tutorials

Like I said, I don't have the Wavelet Toolbox so I can't run your code.

[Suniti Singh]

Ok.. thank you once again

[DGM]

All your incoming images are uint8. Doing all this stuff on a uint8 image will give you a uint8-scale image in floating point, which won't be displayed properly and won't be saved properly because it's improperly-scaled for its class. Blindly normalizing the image will end up in undue brightness and contrast shifts.

Either keep your images properly-scaled for their class, or learn to keep track of class and scale so that you can rescale or recast appropriately when you're done.

% it's uint8-scale, so just cast it to uint8
extracted_watermark = uint8(extracted_watermark);
imshow(extracted_watermark);

I have no idea why you're getting what looks like the cover image out instead of the payload. I don't have WT, and I'm not going to try debugging in the forum editor.

[Suniti Singh]

Sir I worked on it and now extracted the watermark but When I calculated the psnr between input image and watermarked image it gives result as inf.is the embedding and extraction process is incorrect?

[Image Analyst]

You must not be comparing the recovered original image with the watermarked input image. If the PSNR is infininte, you're comparing the image to itself, like either

1. watermarked input image to itself
2. recovered original image estimate to itself
3. watermark (logo) image to itself

If you were able to perfectly separate the watermarked input image into an estimate of the original (not watermarked) image plus the watermark image, then it's possible that the PSNR is infinite. In other words, if your recovery algorithm was so perfect, and then you got the PSNR of the original (not watermarked) image with the estimate (recovered) image, then it would be infinite. You can check that with isequal. Something like

if isequal(perfectNonwatermarkedOriginalImage, recoveredEstimatedImage)
    % Then they're exactly the same pixel for pixel and
    % PSNR would be infinite.
end

[Suniti Singh]

thank you for you reply but sir i am implementing one paper using the same steps an same images with same dimensions and as i said i am getting PSNR value is infinite but the paper i referred had only PSNR values approx to 55dB. If possible can you go through my code to check is there any mistakes in the code it will be a great help for me.

thank you once again for giving your valuable time.

[Image Analyst]

Sorry, I don't have the wavelet toolbox, and I can't really debug or examine variables if I run the code here in answers.

% Step 1: Read the original color image OI of size N x N
OI = imread('mandrill.jpg');
OI = uint8(imresize(OI, [512 512]));

% Step 2: Select the Red color component from the original image
color_component = OI(:, :, 1); % Red color component

% Step 3: Apply DWT to decompose the color component into LL, HL, LH, and HH sub-bands
[LL, HL, LH, HH] = dwt2(color_component, 'haar');

% Step 4: Apply DWT to the LL sub-band to decompose it further
[LL_LL, LL_HL, LL_LH, LL_HH] = dwt2(LL, 'haar');

% Step 5: Apply DCT to the LL_HH sub-band and obtain the DCT coefficient matrix B
B = dct2(LL_HH);

% Step 6: Apply SVD to B and obtain U, S, and V
[U, S, V] = svd(B);

% Step 7: Read the watermark image OW of size N/2 x N/2
OW = imread('peppers.png');
OW = uint8(imresize(OW, [256 256]));
OW = OW(:, :, 1); % Red color component
% Step 8: Apply DWT to decompose the watermark image into WLL, WHL, WLH, and WHH sub-bands
[WLL, WHL, WLH, WHH] = dwt2(OW, 'haar');

% Step 9: Apply DCT to the WHH sub-band and obtain the DCT coefficient matrix D
D = dct2(WHH);

% Step 10: Apply SVD to D and obtain U1, S1, and V1
[U1, S1, V1] = svd(D);

% Step 11: Modify S with the watermark by calculating S2 = S + alpha * S1
alpha = 0.01; % Modify the value of alpha as needed
S2 = S + alpha * S1;

% Step 12: Obtain B* using B* = U * S2 * V'
B_modified = U * S2 * V';

% Step 13: Apply inverse DCT to B* to produce the modified LL_HH* sub-band
LL_HH_modified = idct2(B_modified);

% Step 14: Apply inverse DWT to LL_LL, LL_HL, LL_LH, and LL_HH* to get the modified LL* sub-band
LL_modified = idwt2(LL_LL, LL_HL, LL_LH, LL_HH_modified, 'haar');

% Step 15: Apply inverse DWT to LL*, HL, LH, and HH to get the watermarked image
WI = idwt2(LL_modified, HL, LH, HH, 'haar');

% Step 16: Set the modified color component to the original image
OI_watermarked = OI;
OI_watermarked(:, :, 1) = WI; % Replace the Red component with the watermarked component

% Display and save the watermarked image
figure;
imshow(OI_watermarked);
title('Watermarked Image');

imwrite(OI_watermarked, 'watermarked_image.jpg');
% Step 1: Read the watermarked image
OI_watermarked = imread('watermarked_image.jpg');

% Step 2: Select the Red color component from the watermarked image
color_component_watermarked = OI_watermarked(:, :, 1); % Red color component

% Step 3: Apply DWT to decompose the color component into LL, HL, LH, and HH sub-bands
[LL_watermarked, HL_watermarked, LH_watermarked, HH_watermarked] = dwt2(color_component_watermarked, 'haar');

% Step 4: Apply DWT to the LL sub-band to decompose it further
[LL_LL_watermarked, LL_HL_watermarked, LL_LH_watermarked, LL_HH_watermarked] = dwt2(LL_watermarked, 'haar');

% Step 5: Select the LL_HH* band and apply DCT to it to obtain the DCT coefficient matrix A
A = dct2(LL_HH_watermarked);

% Step 6: Apply SVD to A and obtain WU, WS, and WV
[WU, WS, WV] = svd(A);

% Step 7: Calculate Sr by subtracting the original singular values WS from S
Sr = (S - WS) / alpha;

% Step 8: Obtain Wr using Wr = U1 * Sr * V1'
Wr = U1 * Sr * V1';

% Step 9: Apply inverse DCT to Wr to produce the extracted LL_HH* sub-band
LL_HH_extracted = idct2(Wr);

% Step 10: Apply inverse DWT to LL_LL_watermarked, LL_HL_watermarked, LL_LH_watermarked, and LL_HH_extracted to get the extracted LL* sub-band
LL_extracted = idwt2(LL_LL_watermarked, LL_HL_watermarked, LL_LH_watermarked, LL_HH_extracted, 'haar');

% Step 11: Apply inverse DWT to LL_extracted, HL_watermarked, LH_watermarked, and HH_watermarked to get the extracted watermarked image
OW_extracted = idwt2(LL_extracted, HL_watermarked, LH_watermarked, HH_watermarked, 'haar');

% Set the extracted watermark to the original watermark size
extracted_watermark = imresize(OW_extracted, [256 256]);

% Display the extracted watermark image
figure;
whos extracted_watermark

  Name                       Size              Bytes  Class     Attributes

  extracted_watermark      256x256            524288  double              

imshow(extracted_watermark, []);
title('Extracted Watermark');

% Save the extracted watermark image
imwrite(extracted_watermark, 'extracted_watermark.jpg');