Color image blind watermarking method based on three-dimensional quantization index modulation

Abstract

The invention discloses a color image blind watermarking method based on three-dimensional quantization index modulation. According to the method, image block coefficients are converted into a frequency domain coefficient matrix through Schur decomposition, the maximum characteristic values of three channels serve as three-dimensional carrier signals for overall processing, the three-dimensional signals are quantized through three-dimensional quantization index modulation to complete embedding and blind extraction of watermarks, extraction errors caused by one-dimensional directional offset are avoided, and the extraction accuracy is improved. Therefore, the probability of false extraction of watermarks is reduced, and the robustness is enhanced. The method not only has better robustness, but also has better invisibility, solves the problems of poor robustness and low invisibility of a high-capacity color image digital watermarking algorithm, and is suitable for occasions of high-robustness and high-invisibility high-capacity digital media copyright protection.

Description

Technical Field

[0001] The present invention belongs to the field of information security technology and relates to copyright protection of large-capacity color digital images with strong robustness and high invisibility. Background Art

[0002] With the widespread dissemination and use of digital images in various fields, their value as information carriers has become increasingly prominent. However, this has also brought new challenges to copyright protection. Acts such as illegal copying, tampering, and dissemination not only infringe on the rights of creators but also affect the authenticity and credibility of image content. Digital watermarking technology has therefore become a key means of protecting copyright, combating piracy, and verifying the source of information. An ideal digital watermarking system requires comprehensive optimization in terms of robustness, invisibility, security, and embedding capacity. In recent years, as the requirements for visual representation and information expression of copyright logos continue to increase, color image watermarks have attracted widespread attention due to their higher information carrying capacity and aesthetics. Therefore, developing a color image watermarking algorithm with strong robustness, high invisibility, and large capacity has become a key research direction. Summary of the Invention

[0003] The purpose of the present invention is to provide a color image blind watermarking method based on three-dimensional quantized index modulation, which is characterized by being implemented through a specific watermark embedding process and an extraction process. The watermark embedding process is described as follows: Step 1: First, take a picture of Color digital watermark image Perform dimensionality reduction processing to obtain three layered watermark images , corresponding to the three color channels R, G and B respectively; then, using the key The driven two-dimensional logic adjusts the sine map to encrypt each channel image; next, the single channel image Each pixel value in is converted from decimal to 8-bit binary form, thus generating three single-channel watermark sequences ; Finally, the three single-channel watermark sequences Splicing to get a long watermark sequence ,in, Color digital watermark image The number of edge pixels, Represents the three channels of R, G, and B respectively; Step 2: Make a picture of size Color host image Divide into three layers of non-overlapping pixel blocks of size 2×2, where For color host images The number of edge pixels; Step 3: Use random function to color host image Randomly select Three-layer pixel blocks are used for watermark embedding, and the positions of the selected three-layer pixel blocks are determined by the key Recording, among which, Color digital watermark image The number of edge pixels; Step 4: Select a three-layer pixel block ,right Perform Schur decomposition to obtain a three-level unitary matrix and a three-level upper triangular matrix ,from Select the three largest eigenvalues ,in Represents the three channels of R, G, and B respectively; Step 5: Construct a three-dimensional Cartesian coordinate system, each maximum eigenvalue Corresponding to a coordinate axis direction, a three-dimensional carrier coefficient is defined , according to formula (1) define the embedded reference point ; (1) in, is the rounding function towards zero, is the watermark embedding step size; Step 6: Calculate the three-dimensional carrier coefficient according to formula (2) The normalized residual vector of Position point L in the three-dimensional coordinate system; (2) in, is the watermark embedding step size; Step 7: Define quantization points ,make , , , , , , , ; Divide the quantized points into two mutually exclusive point sets according to the watermark information and ,make , ,in, is the embedded reference point, is the watermark embedding step size; Step 8: From the watermark sequence Read one bit of watermark information to be embedded in sequence ,according to The value of selects the corresponding point set or ; Then, calculate the point The Euclidean distance between all candidate points in the selected point set is used to select the point with the smallest distance as the embedding position. ; According to formula (3), point Mapping back to the spatial domain to obtain the maximum eigenvalue after quantization ; (3) in, is the rounding function towards zero, is the watermark embedding step size, Represents the three channels of R, G, and B respectively; Step 9: The maximum eigenvalue after quantization Updated to , get the upper triangular matrix containing the watermark , and perform inverse Schur decomposition according to formula (4) to obtain the three-layer pixel block containing the watermark ; (4) in, Represents the three channels R, G, and B respectively. express The transpose of Step 10: Repeat steps 4 to 9 until all watermark information is embedded into the three-layer pixel blocks, and finally the watermarked image is obtained. ; The watermark extraction process is described as follows: Step 1: The host image containing watermark information Divide into three layers of non-overlapping pixel blocks of size 2×2, according to the key Extract the information recorded in Three-layer pixel blocks containing watermark information, where Color digital watermark image The number of edge pixels; Step 2: Select a three-layer pixel block containing a watermark ,right Perform Schur decomposition to obtain a three-level unitary matrix and a three-level upper triangular matrix ,from Select the three largest eigenvalues ,in Represents the three channels of R, G, and B respectively; Step 3: Construct a 3D Cartesian coordinate system and define the 3D carrier coefficients , according to formula (5) to define the extraction reference point ; (5) in, is the rounding function towards zero, is the watermark embedding step size; Step 4: Calculate the three-dimensional carrier coefficient according to formula (6) The normalized residual vector of Position point in the three-dimensional coordinate system ,in Represents the three channels of R, G, and B respectively; (6) in, is the watermark embedding step size; Step 5: Calculate the position point using formula (7) Arrival Set and The vertical distance from the nearest point set plane is the watermark information corresponding to the extracted watermark position. ; (7) in, It is a function that calculates the perpendicular distance between a point and a plane; Step 6: Repeat steps 2 to 5 until all watermark information is completely extracted and a complete watermark sequence is obtained. ; Step 7: Extract Equally split to obtain three watermark sequences , each sequence is converted from 8-bit binary to decimal and the key is used Execute the inverse two-dimensional logic adjustment sine mapping algorithm to decrypt and recover three layered watermark images , and finally the layered watermark image Reorganize and finally get the watermark image ,in Represents the three channels R, G, and B respectively. BRIEF DESCRIPTION OF THE DRAWINGS

[0004] Figure 1 It is a size of The original color host image. Figure 2 It is a size of The original color watermark image. Figure 3 It will Figure 2 The watermark image shown is embedded into the host image Figure 1The peak signal-to-noise ratio (PSNR) value of the watermarked image obtained is 40.242dB.

[0005] Figure 4 It is from Figure 3 The normalized cross-correlation coefficient NC value of the watermark extracted from is 1.000.

[0006] Figure 5 (a) Figure 5 (b) Figure 5 (c) Figure 5 (d) Figure 5 (e) Figure 5 (f) Figure 5 (g) Figure 5 (h) Figure 5 (i) Figure 5 (j) is the Figure 3 The watermarks extracted from the watermarked images shown are attacked with JPEG70, JPEG2000 (7:1), salt and pepper noise (1%), Gaussian low-pass filtering (5×5), shearing (1:128; 1:128), scaling (50%), rotation (15°), translation (35, -25), affine (0.01, 0.01), and skew transformation (50, 1; 512, 60; 450, 512; 1, 400). The normalized cross-correlation coefficient (NC) values ​​are 0.986, 0.993, 0.981, 0.984, 0.942, 0.906, 0.984, 0.989, 0.938, and 0.908, respectively. DETAILED DESCRIPTION

[0007] The purpose of the present invention is to provide a color image blind watermarking method based on three-dimensional quantized index modulation, which is characterized by being implemented through a specific watermark embedding process and an extraction process. The watermark embedding process is described as follows: Step 1: First, take a picture of Color digital watermark image Perform dimensionality reduction processing to obtain three layered watermark images , corresponding to the three color channels R, G and B respectively; then, using the key The driven two-dimensional logic adjusts the sine map to encrypt each channel image; next, the single channel image Each pixel value in is converted from decimal to 8-bit binary form (for example, if a pixel value is 168, it is converted into a binary watermark sequence of 10101000), thereby generating three single-channel watermark sequences ; Finally, the three single-channel watermark sequences Splicing to get a long watermark sequence ,in, Color digital watermark image The number of edge pixels, Represents the three channels R, G, and B respectively. ; Step 2: Make a picture of size Color host image Divide into three layers of non-overlapping pixel blocks of size 2×2, where For color host images The number of edge pixels; Step 3: Use random function to color host image Randomly select Three-layer pixel blocks are used for watermark embedding, and the positions of the selected three-layer pixel blocks are determined by the key Recording, among which, Color digital watermark image The number of edge pixels, , ; Step 4: Select a three-layer pixel block ,right Perform Schur decomposition to obtain a three-level unitary matrix and a three-level upper triangular matrix ,from Select the three largest eigenvalues ,in Represent the three channels R, G, and B respectively; here, let ,but , , , , ; Step 5: Construct a three-dimensional Cartesian coordinate system, each maximum eigenvalue Corresponding to a coordinate axis direction, a three-dimensional carrier coefficient is defined , according to formula (1) define the embedded reference point ; (1) in, is the rounding function towards zero, is the watermark embedding step; assuming that, when hour, , , ; Step 6: Calculate the three-dimensional carrier coefficient according to formula (2) The normalized residual vector of Position point L in the three-dimensional coordinate system; (2) in, is the watermark embedding step size; here, when ,but ; Step 7: Define quantization points ,make , , , , , , , ; Divide the quantized points into two mutually exclusive point sets according to the watermark information and ,make , ,in, is the embedded reference point, is the watermark embedding step; at this time, when ,but ; Step 8: From the watermark sequence Read one bit of watermark information to be embedded in sequence ,according to The value of selects the corresponding point set or ; Then, calculate the point The Euclidean distance between all candidate points in the selected point set is used to select the point with the smallest distance as the embedding position. ; According to formula (3), point Mapping back to the spatial domain to obtain the maximum eigenvalue after quantization ; (3) in, is the rounding function towards zero, is the watermark embedding step size, Represents R, G, B three channels respectively; at this time, when , , assuming , then calculate the point and point set The Euclidean distance of all candidate points (E, F, G, H) yields the point with dot The distance is the smallest, so the point As an embed location ;at this time ,but , , ; Step 9: The maximum eigenvalue after quantization Updated to , get the upper triangular matrix containing the watermark , and perform inverse Schur decomposition according to formula (4) to obtain the three-layer pixel block containing the watermark ; (4) in, Represents the three channels R, G, and B respectively. express The transpose of ; at this time, , ; Step 10: Repeat steps 4 to 9 until all watermark information is embedded into the three-layer pixel blocks, and finally the watermarked image is obtained. ; The watermark extraction process is described as follows: Step 1: The host image containing watermark information Divide into three layers of non-overlapping pixel blocks of size 2×2, according to the key Extract the information recorded in Three-layer pixel blocks containing watermark information, where Color digital watermark image The number of edge pixels; Step 2: Select a three-layer pixel block containing a watermark ,right Perform Schur decomposition to obtain a three-level unitary matrix and a three-level upper triangular matrix ,from Select the three largest eigenvalues ,in Represents the three channels of R, G, and B respectively; at this time, , , , , , ; Step 3: Construct a 3D Cartesian coordinate system and define the 3D carrier coefficients , according to formula (5) to define the extraction reference point ; (5) in, is the rounding function towards zero, is the watermark embedding step size; when hour, , , ; Step 4: Calculate the three-dimensional carrier coefficient according to formula (6) The normalized residual vector of Position point in the three-dimensional coordinate system ,in Represents the three channels of R, G, and B respectively; (6) in, is the watermark embedding step; at this time, when When ; Step 5: Calculate the position point using formula (7) Arrival Set and The vertical distance from the nearest point set plane is the watermark information corresponding to the extracted watermark position. ; (7) in, is a function of the vertical distance between the calculation point and the plane; in this case, , , ,therefore ,but , and the watermark information when embedding consistent with the assumptions; Step 6: Repeat steps 2 to 5 until all watermark information is completely extracted and a complete watermark sequence is obtained. ; Step 7: Extract Equally split to obtain three watermark sequences , each sequence is converted from 8-bit binary to decimal and the key is used Execute the inverse two-dimensional logic adjustment sine mapping algorithm to decrypt and recover three layered watermark images , and finally the layered watermark image Reorganize and finally get the watermark image ,in Represents the three channels of R, G, and B respectively; at this time, .

[0008] Verification of the effectiveness of the present invention

[0009] In order to prove the effectiveness of the present invention, Figure 1 The sizes shown are The 24-bit standard image is used as the host image and is used Figure 2 The size shown is The 24-bit color image is used as a digital watermark for verification.

[0010] Figure 3 It will Figure 2 The watermark image shown is embedded into the host image Figure 1 The peak signal-to-noise ratio (PSNR) of the watermarked image obtained is 40.242dB. Figure 4 It is from Figure 3 The normalized cross-correlation coefficient NC value of the watermark extracted from is 1.000; Figure 5 (a) Figure 5 (b) Figure 5 (c) Figure 5 (d) Figure 5 (e) Figure 5 (f) Figure 5 (g) Figure 5 (h) Figure 5 (i) Figure 5 (j) is the Figure 3 The watermarks extracted from the watermarked images shown are attacked with JPEG70, JPEG2000 (7:1), salt and pepper noise (1%), Gaussian low-pass filtering (5×5), shearing (1:128; 1:128), scaling (50%), rotation (15°), translation (35, -25), affine (0.01, 0.01), and skew transformation (50, 1; 512, 60; 450, 512; 1, 400). The normalized cross-correlation coefficient (NC) values ​​are 0.986, 0.993, 0.981, 0.984, 0.942, 0.906, 0.984, 0.989, 0.938, and 0.908, respectively.

[0011] In summary, the color digital watermark images extracted by this method have good identifiability and high NC values ​​under conventional attacks and geometric attacks, indicating that the method has strong robustness; at the same time, the embedded color digital watermark image has good visual effect, which meets the invisibility requirement of the watermark method; in addition, the total running time of this method under Windows11version23H2, AMD Ryzen 5 7500F CPU@3.70GHz, 32GB RAM and MATLAB (R2021b) environment is 1.039 seconds, which meets the high real-time requirements; in addition, the binary information actually embedded in the color digital image is It has a large watermark capacity and meets the needs of digital watermark copyright protection for large-capacity color images.

Claims

1. The purpose of the present invention is to provide a color image blind watermarking method based on three-dimensional quantization index modulation, characterized in that: This is achieved through a specific watermark embedding process and extraction process. The watermark embedding process is described as follows: Step 1: First, take a picture of Color digital watermark image Perform dimensionality reduction processing to obtain three layered watermark images , corresponding to the three color channels R, G and B respectively; then, using the key The driven two-dimensional logic adjusts the sine map to encrypt each channel image; next, the single channel image Each pixel value in is converted from decimal to 8-bit binary form, thus generating three single-channel watermark sequences ; Finally, the three single-channel watermark sequences Splicing to get a long watermark sequence ,in, Color digital watermark image The number of edge pixels, Represents the three channels of R, G, and B respectively; Step 2: Make a picture of size Color host image Divide into three layers of non-overlapping pixel blocks of size 2×2, where For color host images The number of edge pixels; Step 3: Use random function to color host image Randomly select Three-layer pixel blocks are used for watermark embedding, and the positions of the selected three-layer pixel blocks are determined by the key Recording, among which, Color digital watermark image The number of edge pixels; Step 4: Select a three-layer pixel block ,right Perform Schur decomposition to obtain a three-level unitary matrix and a three-level upper triangular matrix ,from Select the three largest eigenvalues ,in Represents the three channels of R, G, and B respectively; Step 5: Construct a three-dimensional Cartesian coordinate system, each maximum eigenvalue Corresponding to a coordinate axis direction, a three-dimensional carrier coefficient is defined , according to formula (1) define the embedded reference point ; (1) in, is the rounding function towards zero, is the watermark embedding step size; Step 6: Calculate the three-dimensional carrier coefficient according to formula (2) The normalized residual vector of Position point L in the three-dimensional coordinate system; (2) in, is the watermark embedding step size; Step 7: Define quantization points ,make , , , , , , , ; Divide the quantized points into two mutually exclusive point sets according to the watermark information and ,make , ,in, is the embedded reference point, is the watermark embedding step size; Step 8: From the watermark sequence Read one bit of watermark information to be embedded in sequence ,according to The value of selects the corresponding point set or ; Then, calculate the point The Euclidean distance between all candidate points in the selected point set is used to select the point with the smallest distance as the embedding position. ; According to formula (3), point Mapping back to the spatial domain to obtain the maximum eigenvalue after quantization ; (3) in, is the rounding function towards zero, is the watermark embedding step size, Represents the three channels of R, G, and B respectively; Step 9: The maximum eigenvalue after quantization Updated to , get the upper triangular matrix containing the watermark , and perform inverse Schur decomposition according to formula (4) to obtain the three-layer pixel block containing the watermark ; (4) in, Represents the three channels R, G, and B respectively. express The transpose of Step 10: Repeat steps 4 to 9 until all watermark information is embedded into the three-layer pixel blocks, and finally the watermarked image is obtained. ; The watermark extraction process is described as follows: Step 1: The host image containing watermark information Divide into three layers of non-overlapping pixel blocks of size 2×2, according to the key Extract the information recorded in Three-layer pixel blocks containing watermark information, where Color digital watermark image The number of edge pixels; Step 2: Select a three-layer pixel block containing a watermark ,right Perform Schur decomposition to obtain a three-level unitary matrix and a three-level upper triangular matrix ,from Select the three largest eigenvalues ,in Represents the three channels of R, G, and B respectively; Step 3: Construct a 3D Cartesian coordinate system and define the 3D carrier coefficients , according to formula (5) to define the extraction reference point ; (5) in, is the rounding function towards zero, is the watermark embedding step size; Step 4: Calculate the three-dimensional carrier coefficient according to formula (6) The normalized residual vector of Position point in the three-dimensional coordinate system ,in Represents the three channels of R, G, and B respectively; (6) in, is the watermark embedding step size; Step 5: Calculate the position point using formula (7) Arrival Set and The vertical distance from the nearest point set plane is the watermark information corresponding to the extracted watermark position. ; (7) in, It is a function that calculates the perpendicular distance between a point and a plane; Step 6: Repeat steps 2 to 5 until all watermark information is completely extracted and a complete watermark sequence is obtained. ; Step 7: Extract Equally split to obtain three watermark sequences , each sequence is converted from 8-bit binary to decimal and the key is used Execute the inverse two-dimensional logic adjustment sine mapping algorithm for decryption and recover the three layered watermark images , and finally the layered watermark image Reorganize and finally get the watermark image ,in Represents the three channels R, G, and B respectively.

Images