A Robust Image Watermarking Algorithm Based on Dual Channels

Abstract

The present invention relates to the field of image processing, and particularly to a robust image watermarking algorithm based on dual channels. After performing DWT on the U channel, subbands are selected for blocking, then DCT is carried out. The values in the watermark sequence are embedded by selecting the DCT mid-frequency coefficients in each block. Then, after performing inverse DCT transformation and then inverse DWT transformation, the U channel image with the embedded watermark is obtained; after performing DWT transformation on the V channel, the selected subbands are subjected to DCT transformation. The DCT coefficients are subjected to SVD to obtain the singular value matrix S. The grayscale image of the second watermark image is subjected to SVD decomposition to obtain the singular value matrix S w , and S w is embedded into the S matrix according to the embedding rules to complete the watermark embedding. Then, inverse SVD decomposition and inverse DCT transformation are carried out to obtain a new wavelet subband, and inverse DWT transformation is performed on it to obtain the V channel image with the embedded watermark. Finally, the image with the embedded watermark is synthesized. The present invention uses different algorithms for watermark image embedding and extraction, making the image with the embedded watermark have stronger robustness and also ensuring inappreciability.

Description

Technical Field

[0001] The present invention relates to the field of image processing, and particularly to a robust image watermarking algorithm based on dual channels. Background Art

[0002] With the vigorous development of information technology and the further improvement of people's copyright awareness, digital image watermarking has been more and more widely used in the field of copyright protection of digital images.

[0003] Most of the existing image watermarking technologies use a single image watermarking algorithm for watermark embedding. Once an attack method with poor robustness to the watermarking algorithm is used to attack the watermarked image, the extracted watermarked image will be seriously distorted. Therefore, the present invention has made improvements in this regard. Two image watermarking algorithms with different robustness are used for watermark embedding on two channels of the original carrier image, which can improve the robustness of the watermark and at the same time ensure the imperceptibility of the watermark. Summary of the Invention

[0004] The object of the present invention is to improve the robustness of image watermarking, and thus a robust image watermarking algorithm based on dual channels is provided.

[0005] To achieve the above object, the present invention adopts the following technical solutions:

[0006] A robust image watermarking algorithm based on dual channels, including watermark embedding and watermark extraction, and the specific steps are as follows:

[0007] Step1: Preprocessing of the carrier image and the watermark image;

[0008] Convert the carrier image in RGB format to YUV format, select the U channel as the embedding channel for the first watermark image, and select the V channel as the embedding channel for the second watermark image; read the grayscale image W1 of the first watermark image and the grayscale image W2 of the second watermark image, and convert W1 into a one-dimensional watermark sequence;

[0009] Step2: Embedding the watermark in the U channel;

[0010] Perform discrete wavelet transform (DWT) on the U channel of the carrier image to obtain the selected wavelet subbands, perform discrete cosine transform (DCT) after dividing the wavelet subbands into blocks, select the DCT coefficients and embed the watermark sequence values of W1 therein according to the embedding rules, and perform inverse DCT transform and inverse DWT transform on the divided sub-image blocks to obtain the U channel image with the embedded watermark;

[0011] Step3: Embedding the watermark in the V channel;

[0012] Perform DWT transformation on the V channel of the carrier image to obtain the selected wavelet subband, and perform DCT transformation on the wavelet subband; perform singular value decomposition (SVD) on the DCT coefficients, and at the same time perform SVD decomposition on W2. The singular value matrices obtained from the two decompositions are used for watermark embedding using the embedding rule to obtain a new singular value matrix; perform inverse SVD decomposition and inverse DCT transformation on the new singular value matrix to obtain a new wavelet subband with the watermark embedded. After performing inverse DWT transformation on the new wavelet subband, obtain the V channel image with the watermark embedded;

[0013] Step4: Combine channels to obtain the image with the watermark embedded;

[0014] Perform channel combination on the U channel with the watermark embedded, the V channel with the watermark embedded, and the Y channel of the carrier picture without the watermark embedded, and then convert it to the RGB format to obtain the image with the watermark embedded;

[0015] Step5: Extract the watermark in the U channel;

[0016] Perform DWT transformation on the image with the watermark embedded and the U channel of the carrier picture to select the wavelet subband. After dividing the wavelet subbands of both into blocks, perform DCT transformation; use the inverse operation of the embedding rule to extract the watermark sequence value from the DCT coefficients of the image with the watermark embedded; then reconstruct the watermark sequence composed of the watermark sequence values into a watermark image to obtain the extracted watermark image W1';

[0017] Step6: Extract the watermark in the V channel;

[0018] Perform DWT transformation on the image with the watermark embedded and the V channel of the original image to select the wavelet subband, perform DCT transformation on the wavelet subbands of both, and perform SVD decomposition on the DCT coefficients to obtain a singular value matrix; use the inverse operation of the embedding rule to extract the singular value matrix of the embedded watermark from the two singular value matrices; finally, perform inverse SVD decomposition on the extracted singular value matrix to obtain the extracted watermark image W2'.

[0019] In Step1, the size of the carrier picture is M×M, the size of the first watermark image grayscale map W1 is N1×N1, N1 = M / 16, and the size of the second watermark image grayscale map W2 is N2×N2, N2 = M / 4.

[0020] The specific steps of Step2 are as follows:

[0021] Step2.1: Perform one-level DWT transformation on the U channel of the carrier image using the haar wavelet, and the selected subband is the low-frequency approximation subband LL;

[0022] Step2.2: Divide the LL subband into 8×8 blocks, perform DCT transformation on each sub-image block, and select 8 DCT intermediate-frequency coefficients for embedding, using x wi = xi +ew j The rule embeds the watermark sequence value of W1; where x i represents the DCT intermediate frequency coefficient, i represents the i-th intermediate frequency coefficient, e represents the embedding strength, w j represents the value of the watermark sequence, j represents the j-th watermark sequence value, x wi represents the DCT intermediate frequency coefficient after watermark embedding; finally, the inverse DCT transform is performed on the sub-image block with embedded watermark to obtain the sub-band LL w , and then the inverse DWT transform is performed to obtain the U-channel image with embedded watermark.

[0023] The specific steps of the said Step3 are as follows:

[0024] Step3.1: Perform a two-level DWT transform on the V channel of the carrier image using the haar wavelet, and select the sub-band as the low-frequency approximation sub-band LL2;

[0025] Step3.2: Perform a DCT transform and an SVD decomposition on LL2 to obtain the singular value matrix S. At the same time, perform an SVD decomposition on W2 to obtain the singular value matrix S w , and adopt the rule of S' = S + αS w for watermark embedding, where α is the embedding strength, S' is the singular value matrix after watermark embedding. After inverse SVD decomposition of S', perform an inverse DCT transform to obtain the sub-band LL 2w ; finally, perform an inverse DWT transform on LL 2w to obtain the V-channel image with embedded watermark.

[0026] The specific steps of the said Step5 are as follows:

[0027] Step5.1: Perform a one-level DWT transform on the U channels of the watermarked image and the carrier image using the haar wavelet to obtain the sub-bands LL w and LL. Divide LL w and LL into 8×8 blocks and perform a DCT transform on each block;

[0028] Step5.2: Extract the DCT intermediate frequency coefficients at the corresponding embedding positions of the two, and extract the watermark sequence value according to the rule of w' j =(x wi -x i ) / e. Where w' j is the extracted watermark sequence value. The watermark sequence values are combined into a watermark sequence for reconstruction to obtain the extracted watermark image W1'.

[0029] The specific steps of the said Step6 are as follows:

[0030] Step6.1: Perform a two - level DWT transform on the V channels of the watermarked embedded image and the carrier image using the haar wavelet to obtain the sub - bands LL 2w and LL2, and perform a DCT transform on LL 2w and LL;

[0031] Step6.2: Perform an SVD decomposition on the DCT coefficients to obtain the singular value matrices S' and S, and extract the watermark according to the rule S' w =(S' - S) / α, where S' w is the singular value matrix for extracting the watermark. Perform an inverse SVD decomposition on S' w to obtain the extracted watermarked image W2'.

[0032] Beneficial effects:

[0033] The dual - channel image watermarking method provided by the present invention embeds the watermark using two robust watermarking algorithms based on the U channel and the V channel of the carrier image, making the watermarked image more robust against different attacks, and at the same time ensuring the imperceptibility of the watermarked embedded image. Brief description of the drawings

[0034] Attached Figure 1 is the watermark embedding process diagram of the present invention.

[0035] Attached Figure 2 is the watermark extraction process diagram of the present invention.

[0036] Attached Figure 3 are the original carrier image and the watermark image of the present invention.

[0037] Attached Figure 4 are the carrier image with the watermark embedded without attack and the extracted watermark image of the present invention.

[0038] Attached Figure 5 are the carrier image with the watermark embedded under exposure attack and the extracted watermark image of the present invention.

[0039] Attached Figure 6 are the carrier image with the watermark embedded under saturation attack and the extracted watermark image of the present invention.

[0040] Attached Figure 7 are the carrier image with the watermark embedded under salt - and - pepper noise attack and the extracted watermark image of the present invention.

[0041] Attached Figure 8 are the carrier image with the watermark embedded under Gaussian noise attack and the extracted watermark image of the present invention.

[0042] Attached Figure 9 are the carrier image with the watermark embedded under median filtering attack and the extracted watermark image of the present invention.

[0043] Appendix Figure 10 These are the carrier image after embedding the watermark by the mean filtering attack of the present invention and the extracted watermark image.

[0044] Appendix Figure 11 These are the carrier image after embedding the watermark by the cropping attack of the present invention and the extracted watermark image.

[0045] Appendix Figure 12 These are the carrier image after embedding the watermark by the rotation attack of the present invention and the extracted watermark image.

[0046] Appendix Figure 13 These are the carrier image after embedding the watermark by the JPEG compression attack of the present invention and the extracted watermark image. Detailed implementation manners

[0047] In order to make the technical solutions to be solved by the present invention clearer and more understandable, the present invention will be described in detail below with reference to the accompanying drawings and specific embodiments. It should be noted that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0048] Embodiment 1

[0049] As Figure 1-2 shown, a robust image watermarking algorithm based on dual channels includes watermark embedding and watermark extraction, and the specific steps are as follows:

[0050] Step1 Preprocessing of the carrier image and the watermark image

[0051] Convert the carrier image in RGB format to YUV format, select the U channel as the embedding channel for the first watermark image, and select the V channel as the embedding channel for the second watermark image; read the grayscale image W1 of the first watermark image and the grayscale image W2 of the second watermark image, and convert W1 into a one-dimensional watermark sequence;

[0052] Step2 Embedding the watermark in the U channel

[0053] Perform DWT transformation on the U channel of the carrier image to obtain the selected wavelet subbands, perform DCT transformation on the wavelet subbands after partitioning, select the DCT coefficients and embed the watermark sequence values of W1 therein, and perform inverse DCT transformation and inverse DWT transformation on the partitioned sub-image blocks to obtain the U channel image with the embedded watermark;

[0054] Step3 Embedding the watermark in the V channel

[0055] Perform DWT transformation on the V channel of the carrier image to obtain the selected wavelet subband, and perform DCT transformation on the wavelet subband; perform SVD decomposition on the DCT coefficients, and at the same time perform SVD decomposition on W2. The singular value matrices obtained from the two decompositions are used to embed the watermark using the embedding rule to obtain a new singular value matrix; perform inverse SVD decomposition and inverse DCT transformation on the new singular value matrix to obtain a new wavelet subband with the watermark embedded. After performing inverse DWT transformation on the new wavelet subband, obtain the V channel image with the watermark embedded;

[0056] Step4 Channel merging to obtain the watermark-embedded image

[0057] Merge the U channel with the watermark embedded, the V channel with the watermark embedded, and the Y channel of the carrier picture without the watermark embedded, and then convert it to the RGB format to obtain the watermark-embedded image;

[0058] Step5 Watermark extraction from the U channel

[0059] Perform DWT transformation on the watermark-embedded image and the U channel of the carrier picture to select the wavelet subband. After dividing the wavelet subbands of both into blocks, perform DCT transformation; use the inverse operation of the embedding rule to extract the watermark sequence values from the DCT coefficients of the watermark-embedded image; then reconstruct the watermark sequence composed of the watermark sequence values into a watermark image to obtain the extracted watermark image W1';

[0060] Step6 Watermark extraction from the V channel

[0061] Perform DWT transformation on the watermark-embedded image and the V channel of the original image to select the wavelet subband, perform DCT transformation on the wavelet subbands of both, and perform SVD decomposition on the DCT coefficients to obtain a singular value matrix; use the inverse operation of the embedding rule to extract the singular value matrix with the watermark embedded from the two singular value matrices; finally, perform inverse SVD decomposition on the extracted singular value matrix to obtain the extracted watermark image W2'.

[0062] In Step1, the size of the carrier picture is M×M, the size of the first watermark image grayscale map W1 is N1×N1, N1 = M / 16, and the size of the second watermark image grayscale map W2 is N2×N2, N2 = M / 4.

[0063] In Step2, the embedding method is to use x wi =x i +ew j to embed the regular watermark sequence values in the DCT intermediate frequency coefficients, where x i represents the selected DCT intermediate frequency coefficient, i represents the i-th intermediate frequency coefficient, e represents the embedding strength, w j represents the value of the watermark sequence, and j represents the j-th watermark sequence value.

[0064] In Step 3, the method of embedding is to use the rule of S' = S + αS for the singular value matrix after SVD decomposition, where S is the singular value matrix of the carrier image, S w is the singular value matrix of the watermark image, and α is the embedding strength. w For the watermark extraction in Step 5, the extraction method adopted is to extract the watermark sequence values embedded in the DCT middle frequency coefficients according to the rule of w'

[0065] =(x j -x wi ) / e, where w' i is the extracted watermark sequence value, j represents the j-th watermark sequence value, x j is the middle frequency coefficient embedded with the watermark, x wi is the DCT middle frequency coefficient of the original carrier image, i represents the i-th middle frequency coefficient, and e is the embedding strength. i

[0066] In Step 6, the extraction method adopted is to extract the watermark image in the U channel from the singular value matrix after SVD decomposition according to the rule of S' w =(S' - S) / α, where S' w is the singular value matrix for extracting the watermark, S' is the singular value matrix of the embedded watermark image, S is the singular value matrix of the original carrier image, and α is the embedding strength.

[0067] Embodiment 2

[0068] To verify the effectiveness of the present invention, in this embodiment, pycharm is used as the experimental simulation platform. The size of the original carrier image is 1024×1024, the size of the first watermark image grayscale map W1 embedded in the U channel is 64×64, the Haar wavelet basis function is used for wavelet transform, and the embedding strength is 0.1. The size of the second watermark image grayscale map W2 embedded in the V channel is 256×256, the Haar wavelet basis function is used for wavelet transform, and the embedding strength is 0.08.

[0069] The specific process is as follows:

[0070] Watermark embedding in the U channel

[0071] Perform a one-level DWT transform on the U channel of the carrier image using the Haar wavelet to obtain four sub-bands: LL, LH, HL, and HH. The size of the carrier image is 1024×1024, and the size of the sub-bands is 512×512.

[0072] Select the LL sub-band and divide it into 8×8 sub-image blocks to obtain 64×64 image blocks.

[0073] ​The sub-image blocks are sequentially subjected to DCT transformation, and 64 DCT coefficients are obtained for each sub-block. The size of the grayscale image W1 of the first watermark image is 64×64, and the watermark sequence obtained by its conversion is embedded in 8 DCT intermediate frequency coefficients x selected from the image blocks. i Embedding is performed.

[0074] The inverse DCT transformation is performed on each sub-block embedded with the watermark to obtain a new sub-band LL embedded with the watermark. w 。

[0075] LL w is wavelet reconstructed with the LH, HL, and HH sub-bands to obtain the U-channel image embedded with the watermark.

[0076] V-channel watermark embedding

[0077] The V-channel of the carrier image is subjected to two-level DWT transformation using the haar wavelet to obtain seven sub-bands: LL2, LH2, HL2, HH2, LH, HL, and HH. The size of the carrier image is 1024×1024, and the size of the LL2 sub-band is 256×256.

[0078] After selecting the LL2 sub-band for DCT transformation, the DCT coefficient matrix X is subjected to SVD decomposition to obtain the singular value matrix S.

[0079] U×S×(V) T =SVD(X)

[0080] The grayscale image W2 of the watermark image read is subjected to SVD decomposition to obtain the singular value matrix S w 。

[0081] U W ×S w ×(V W ) T =SVD(W2)

[0082] S w is embedded in S using the following formula to obtain S', where α is the embedding strength with a value of 0.08.

[0083] S'=S+αS w

[0084] The inverse SVD decomposition is performed on the matrix S' to obtain the matrix DCT coefficient matrix X w 。

[0085] X w =U×S'×(V) T

[0086] X w is inverse DCT transformed to obtain the sub-band LL embedded with the watermark. 2w 。

[0087] Inverse DWT transform is performed on the LL 2w sub-band and LH2, HL2, HH2, LH, HL, HH to obtain the V-channel image with the embedded watermark.

[0088] U-channel watermark extraction

[0089] Perform wavelet transform on the U-channel of the watermarked image and the original carrier image to obtain LL, LH, HL, HH and LL w , LH w , HL w , HH w sub-bands.

[0090] Perform 8×8 block partitioning on the LL and LL w sub-bands, and extract the intermediate frequency coefficients at the watermark embedding positions of the two sub-blocks to obtain the embedded watermark sequence value w' using the following formula j , where e is the embedding strength with a value of 0.1.

[0091] w' j = (x wi - x i ) / e

[0092] Combine the watermark sequence values w' j to obtain the watermark sequence w', and reconstruct it to obtain the extracted watermark image W1'.

[0093] V-channel watermark extraction

[0094] Perform DWT transform on the V-channel of the original carrier image and the watermarked image to obtain LL2, LH2, HL2, HH2, LH, HL, HH and LL 2w , LH 2w , HL 2w , HH 2w , LH w , HL w , HH w sub-bands.

[0095] Select the LL2 and LL 2w sub-bands, perform DCT transform on them, and then perform SVD decomposition on the DCT coefficients to obtain the singular value matrices S and S'.

[0096] U×S×(V) T = SVD(LL2)

[0097] U'×S'×(V') T = SVD(LL 2w )

[0098] Obtain the singular value matrix S' from the S and S' matrices using the following formula w, α is the embedding strength, and its value is 0.08.

[0099] S' w =(S' - S) / α

[0100] After performing inverse SVD decomposition on S' w with U w and V w the extracted watermark image W2' is obtained.

[0101] W2' = U w ×S' w ×(V w ) T

[0102] As Figure 3 shown, Host image is the original carrier image, Watermark image1 is the watermark image embedded in the U channel, and Watermark image2 is the watermark image embedded in the V channel. Figure 4 In it, Watermarked image is the carrier image after embedding the watermark without attack, Extratedwatermark1 is the watermark image extracted from the U channel without attack, and Extratedwatermark2 is the watermark image extracted from the V channel without attack.

[0103] The present invention uses the peak signal-to-noise ratio (PSNR) and the structural similarity (SSIM) to reflect the imperceptibility. The higher the peak signal-to-noise ratio (PSNR) of the carrier picture after embedding the watermark, the better the visual quality of the image. Generally, when the peak signal-to-noise ratio (PSNR) is above 30 dB, subjectively, the human vision cannot perceive an obvious change in the image; the structural similarity (SSIM) is used to reflect the similarity degree of the direct structure between the original carrier picture and the carrier picture after embedding the watermark. The value of the structural similarity (SSIM) is 0 - 1. When the calculated value of the structural similarity of the carrier picture after embedding the watermark is closer to 1, it means that it is more similar to the original carrier picture, that is, the imperceptibility is better.

[0104] The present invention uses the normalized correlation coefficient (NC) to reflect the robustness. The value of the normalized correlation coefficient (NC) is 0 - 1. When the calculated value of the normalized correlation coefficient (NC) of the extracted watermark picture is closer to 1, it means that the extracted watermark picture is more similar to the original watermark picture, that is, the robustness is better.

[0105] After experiments, as Figures 5-13 shown, are the carrier images after embedding the watermark and the extracted watermark images under different types of attacks. The specific experimental results are shown in Table 1.

[0106] Table 1 shows the peak signal-to-noise ratio (PSNR), structural similarity (SSIM) of the carrier image embedded with watermark when suffering from different types of attacks, and the normalized correlation coefficient (NC) between the extracted watermark and the original watermark after suffering from different types of attacks. Here, NC1 represents the NC value calculated from the extracted watermark image W1', and NC2 represents the NC value calculated from the extracted watermark image W2'.

[0107] Table 1 Comparison table of various indicators after different types of attacks

[0108]

[0109] It can be found from Table 1 that for the image watermark algorithm extracted by the present invention when dealing with common attacks, the watermarks embedded in the two channels have different effects. At least one of the two watermarks has good robustness against this type of attack. At the same time, it also shows good peak signal-to-noise ratio when there is no attack, and has good imperceptibility.

Claims

1. A robust image watermarking algorithm based on dual channels, characterized in that, It includes watermark embedding and watermark extraction, and the specific steps are as follows: Step1: Preprocess the carrier image and the watermark image; Convert the carrier image in RGB format to YUV format, select the U channel as the embedding channel for the first watermark image, and select the V channel as the embedding channel for the second watermark image; Read the grayscale image W1 of the first watermark image and the grayscale image W2 of the second watermark image, and convert W1 into a one-dimensional watermark sequence; Step2: Embed the watermark in the U channel; Perform DWT transformation on the U channel of the carrier image to obtain the selected wavelet subband. After dividing the wavelet subband into blocks, perform DCT transformation. Select the DCT coefficients and embed the watermark sequence values of W1 into them using the embedding rule. Perform inverse DCT transformation and inverse DWT transformation on the divided sub-image blocks to obtain the U channel image with the embedded watermark; Step3: Embed the watermark in the V channel; Perform DWT transformation on the V channel of the carrier image to obtain the selected wavelet subband, and perform DCT transformation on the wavelet subband; Decompose the DCT coefficients by SVD, and at the same time decompose W2 by SVD. The singular value matrices obtained by the two decompositions are used for watermark embedding using the embedding rule to obtain a new singular value matrix; Perform inverse SVD decomposition and inverse DCT transformation on the new singular value matrix to obtain a new wavelet subband with the embedded watermark. After performing inverse DWT transformation on the new wavelet subband, obtain the V channel image with the embedded watermark; Step4: Merge the channels to obtain the image with the embedded watermark; Merge the U channel with the embedded watermark, the V channel with the embedded watermark, and the Y channel of the carrier image without the embedded watermark, and then convert it to RGB format to obtain the image with the embedded watermark; Step5: Extract the watermark in the U channel; Perform DWT transformation on the image with the embedded watermark and the U channel of the carrier image to select the wavelet subband. After dividing the wavelet subbands of both into blocks, perform DCT transformation; Use the inverse operation of the embedding rule to extract the watermark sequence values from the DCT coefficients of the embedded watermark; Then reconstruct the watermark sequence composed of the watermark sequence values into a watermark image to obtain the extracted watermark image W1'; Step6: Extract the watermark in the V channel; Perform DWT transformation on the image with the embedded watermark and the V channel of the original image to select the wavelet subband, and perform DCT transformation on the wavelet subbands of both. Decompose the DCT coefficients by SVD to obtain a singular value matrix; Use the inverse operation of the embedding rule to extract the singular value matrix of the embedded watermark from the two singular value matrices; Finally, perform inverse SVD decomposition on the extracted singular value matrix to obtain the extracted watermark image W2'; The specific steps of Step2 are as follows: Step2.1: Perform one-level DWT transformation on the U channel of the carrier image using the haar wavelet, and the selected subband is the low-frequency approximation subband LL; Step 2.2: Divide the LL sub-band into 8×8 blocks. After performing DCT transformation on each sub-image block, select 8 DCT mid-frequency coefficients for embedding, and use the rule of x wi = x i + ew j to embed the watermark sequence values of W1; where x i represents the DCT mid-frequency coefficient, i represents the i-th mid-frequency coefficient, e represents the embedding strength, w j represents the value of the watermark sequence, j represents the j-th watermark sequence value, and x wi represents the DCT mid-frequency coefficient after embedding the watermark; finally, perform inverse DCT transformation on the sub-image block embedded with the watermark to obtain the sub-band LL w , and then perform inverse DWT transformation to obtain the U-channel image embedded with the watermark; The specific steps of Step3 are as follows: Step3.1: Perform two-level DWT transformation on the V channel of the carrier image using the haar wavelet, and the selected subband is the low-frequency approximation subband LL2; Step 3.2: Perform DCT transformation and SVD decomposition on LL2 to obtain the singular value matrix S. At the same time, perform SVD decomposition on W2 to obtain the singular value matrix S w , and use the rule of S' = S + αS w for watermark embedding, where α is the embedding strength, S' is the singular value matrix after watermark embedding. After inverse SVD decomposition of S', perform inverse DCT transformation to obtain the sub-band LL with the embedded watermark 2w ; Finally, perform inverse DWT transformation on LL 2w to obtain the V-channel image with the embedded watermark.

2. The robust image watermarking algorithm based on dual channels according to claim 1, wherein In Step1, the size of the carrier image is M×M, the size of the grayscale image W1 of the first watermark image is N1×N1, N1 = M / 16, and the size of the grayscale image W2 of the second watermark image is N2×N2, N2 = M / 4.

3. The robust image watermarking algorithm based on dual channels according to claim 1, wherein The specific steps of Step 5 are as follows: Step5.1: Perform a first-level DWT transformation on the U channels of the watermarked embedded image and the carrier image using the haar wavelet to obtain the subbands LL w and LL, and divide LL w and LL into 8×8 blocks and perform DCT transformation on each block; Step5.2: Extract the DCT mid-frequency coefficients at the corresponding embedding positions of the two, and extract the watermark sequence values according to the rule of w' j =(x wi -x i ) / e, where w' j is the extracted watermark sequence value. Combine the watermark sequence values to form a watermark sequence and reconstruct it to obtain the extracted watermark image W1'.

4. The robust image watermarking algorithm based on dual channels according to claim 1, wherein, The specific steps of Step 6 are as follows: Step6.1: Perform a two-level DWT transformation on the V channel of the watermarked embedded image and the carrier image using the haar wavelet to obtain the subbands LL 2w and LL2, and perform a DCT transformation on LL 2w and LL; Step 6.2: Perform SVD decomposition on the DCT coefficients to obtain the singular value matrices S' and S, and perform watermark extraction according to the rule of S' w =(S' - S) / α, where S' w is the singular value matrix for watermark extraction. Perform inverse SVD decomposition on S' w to obtain the extracted watermark image W2'.

Images