Color image blind watermarking method based on double-axis collaborative quantization index modulation

Abstract

Along with wide propagation of digital media, the copyright protection problem is increasingly prominent. In order to effectively protect the copyright of a digital image, the invention discloses a color image blind watermarking method based on double-axis collaborative quantization index modulation. The method comprises the following steps: firstly, performing high-energy aggregation on image blocks through a reversible matrix fusion domain formula; on the basis, high-energy coefficients of left and right side areas of an image block are synchronously modulated through a double-axis cooperation mechanism, and double-axis cooperation quantization index modulation is realized through a linkage structure, so that embedding and blind extraction of the watermark are completed. And meanwhile, the watermark is preprocessed by adopting a chaotic encryption technology before embedding, so that the security and robustness of the watermark are further enhanced. According to the method, while the invisibility of the watermark is ensured, the capability of resisting common attacks is remarkably improved, the calculation process is simplified, and the method is suitable for carrying out rapid and safe copyright protection on the color image.

Description

Technical Field

[0001] This invention belongs to the field of information security technology and relates to rapid copyright protection for large-capacity color digital images. Background Technology

[0002] With the rapid development of the digital age, the creation, dissemination, and consumption of digital media have undergone tremendous changes. Traditional copyright protection methods are proving inadequate in the face of the ease with which digital content can be copied and tampered with. Therefore, developing efficient and reliable digital watermarking technology has become crucial for protecting digital media copyright. However, traditional one-dimensional quantization index modulation methods only quantize and modulate single feature coefficients in an image, making them less robust to attacks such as compression, noise, and rotation. Furthermore, existing methods are mostly designed for grayscale images, failing to adequately utilize the multi-channel characteristics of color images, thus limiting their practical application in color media copyright protection. The color image blind watermarking method based on dual-axis collaborative quantization index modulation was invented to meet these needs. Summary of the Invention

[0003] The purpose of this invention is to provide a blind watermarking method for color images based on dual-axis collaborative quantization index modulation. Its key feature is that it is achieved through specific watermark embedding and extraction processes, the watermark embedding process of which is described below: Step 1: Convert an image with a pixel size of 24-bit color watermarked image Layered watermark image consisting of three single color channels: red, green, and blue. Then, each single color channel is sequentially... Perform based on 4 keys , , , The four-dimensional Lorenz chaotic map is encrypted, and each encrypted decimal pixel value is converted into an 8-bit binary sequence, then these 8-bit binary sequences are converted into a 4-bit quaternary sequence; then, The quaternary sequences in the sequence are concatenated to form a length of . string sequence ,in, For color watermarked images The number of edge pixels, These represent the red, green, and blue channels, respectively. Step 2: Convert an image with a pixel size of... 24-bit color host image Layered host image divided into three single color channels: red, green, and blue. Then, each layer of host image of The region is divided into areas of size Non-overlapping image patches, where, For color host image The number of edge pixels, These represent the red, green, and blue channels, respectively. Step 3: In the layered host image Select an image patch sequentially from top to bottom and from left to right. Divide it into sizes of left image patch and size are right-side image patch Based on the derived formulas (1) and (2) for the fusion field of invertible matrices, respectively... and Achieving high-energy focusing yields biaxial high-energy coefficients and ,in, These represent the red, green, and blue channels, respectively. (1) (2) in, Represents image blocks No. Line number Column pixel values, Represents image blocks No. Line number The pixel values ​​of the column; Step 4: Calculate the biaxial high-energy coefficients according to formulas (3) and (4). and remainder and ; (3) (4) in, It is a modulo function. For the quantization step size of different channels, , This serves as the global baseline step size. These represent the red, green, and blue channels, respectively. Step 5: Define the embedded quantization point according to formula (5). ; (5) Step 6: From the layered watermark bit sequence In the process, extract one watermark information to be embedded in sequence. According to the embedded watermark information At the defined embedding quantization point Select the watermark information The 8 unique corresponding quantization points; calculate the Euclidean distance according to formula (6) and return the position index of the quantization point corresponding to the shortest Euclidean distance. ; (6) in, For the quantization step size of different channels, , This serves as the global baseline step size. for coordinates in For the position index of the coordinates, An index function that returns the minimum value. These represent the red, green, and blue channels, respectively. Step 7: Obtain the coordinates of the nearest quantization point according to formula (7). As a target quantification point; (7) Step 8: According to formulas (8) and (9), respectively... and High energy coefficients with watermarking were obtained by implementing dual-axis collaborative quantization index modulation. and ; (8) (9) The meanings of the variables in formulas (8) and (9) are as described above; Step 9: Based on the derived formulas (10) and (11) for allocating pixel changes in the invertible matrix fusion domain, the watermarked floating-point image block is obtained. and ; (10) (11) Step 10: Convert the watermarked floating-point image block matrix and Rounding yields an integer image patch matrix containing the watermark. and ,Will and merged into of And update it to its hierarchical host image. The corresponding position in, where, These represent the red, green, and blue channels, respectively. Step 11: Repeat steps 3 through 10 of this process until all watermark information has been embedded, thus obtaining a watermarked layered host image. And the layered host image containing the watermark's three red, green, and blue single-color channels Reassemble and obtain pixel size Watermarked images ,in, These represent the red, green, and blue channels, respectively. The watermark extraction process is described as follows: Step 1: Convert an image with a pixel size of 24-bit color watermarked image A layered watermarked image consisting of three single color channels: red, green, and blue. Then, each layer of watermarked image... of The region is divided into areas of size Non-overlapping watermarked image blocks; among which, These represent the red, green, and blue channels, respectively. Step 2: Layering the watermarked image Select a watermarked image block in the order from top to bottom and from left to right. Divide it into sizes of The watermarked image block on the left and size are The watermarked image block on the right Based on the derived formulas (12) and (13) for the fusion field of invertible matrices, respectively... and Achieving high-energy focusing yields biaxial high-energy coefficients and ; (12) (13) in, Represents image blocks No. Line number Column pixel values, Represents watermarked image blocks No. Line number The pixel values ​​of the column; Step 3: Calculate the biaxial high energy coefficients according to formulas (14) and (15) respectively. and remainder and ; (14) (15) in, It is a modulo function. For the quantization step size of different channels, , This serves as the global baseline step size. These represent the red, green, and blue channels, respectively. Step 4: Define the extraction quantization points according to formula (16). ; (16) Step 5: Calculate the Euclidean distance according to formula (17) and return the position index of the quantized point corresponding to the shortest Euclidean distance. ; (17) in, For the quantization step size of different channels, , This serves as the global baseline step size. for coordinates For the position index of the coordinates, An index function that returns the minimum value. These represent the red, green, and blue channels, respectively. Step 6: Extract watermark information according to formula (18) ; (18) Step 7: Repeat steps 2 through 6 of this process to extract the quaternary watermark bit sequence for each layer. Then, each group of 4 quaternary bits is converted into an 8-bit binary sequence, and then into a decimal pixel value. These represent the red, green, and blue channels, respectively. Step 8: Perform a four-key-based process on each layer of decimal pixels after conversion. , , , Decrypting the four-dimensional Lorenz chaotic map and obtaining layered watermarks ,in, These represent the red, green, and blue channels, respectively. Step 9: Combine layered watermarks Forming the final watermark extraction ,in, These represent the red, green, and blue channels, respectively. Attached Figure Description

[0004] Figure 1 (a) Figure 1 (b) are two original color host images.

[0005] Figure 2 (a) Figure 2 (b) are two original color watermark images.

[0006] Figure 3 (a) Figure 3 (b) is to Figure 2 (a) Figure 2 (b) The watermarks shown are embedded into the host image respectively. Figure 1 (a) Figure 1 The structural similarity (SSIM) values ​​of the watermarked images obtained after (b) are 0.9835 and 0.9788, respectively, and their peak signal-to-noise ratio (PSNR) values ​​are 41.1188 dB and 41.1552 dB, respectively.

[0007] Figure 4 (a) Figure 4 (b) are respectively from Figure 3 (a) Figure 3 The watermarks extracted in (b) have normalized cross-correlation coefficient (NC) values ​​of 1.0000 and 1.0000, respectively.

[0008] Figure 5 (a) Figure 5 (b) Figure 5 (c) Figure 5 (d) Figure 5 (e) Figure 5 (f) is to Figure 3 The watermarked images shown in (a) were subjected to JPEG compression (70), JPEG2000 (4:1), salt and pepper noise (0.2%), scaling (400%), rotation (5°), and cropping (12.5%) attacks, and the normalized cross-correlation coefficients (NC) of the extracted watermarked images were 0.9962, 0.9975, 0.9855, 0.9996, 0.9801, and 0.9394, respectively.

[0009] Figure 6 (a) Figure 6 (b) Figure 6 (c) Figure 6 (d) Figure 6 (e) Figure 6 (f) is to Figure 3The watermarked images shown in (b) were subjected to JPEG compression (70), JPEG2000 (4:1), salt and pepper noise (0.2%), scaling (400%), rotation (5°), and cropping (12.5%) attacks, and the normalized cross-correlation coefficients (NC) of the extracted watermarked images were 0.9978, 1.0000, 0.9853, 1.0000, 0.9854, and 0.9053, respectively. Detailed Implementation

[0010] The purpose of this invention is to provide a blind watermarking method for color images based on dual-axis collaborative quantization index modulation. Its key feature is that it is achieved through specific watermark embedding and extraction processes, the watermark embedding process of which is described below: Step 1: Convert an image with a pixel size of 24-bit color watermarked image Layered watermark image consisting of three single color channels: red, green, and blue. Then, each single color channel is sequentially... Perform based on 4 keys , , , The four-dimensional Lorenz chaotic map is encrypted, and each encrypted decimal pixel value is converted into an 8-bit binary sequence (for example, the decimal pixel value 201 is converted into the binary value 11001001, and then into the quaternary value 3021). Then, this 8-bit binary sequence is converted into a 4-bit quaternary sequence. Then... The quaternary sequences in the sequence are concatenated to form a length of . string sequence ,in, For color watermarked images The number of edge pixels, These represent the red, green, and blue channels, respectively. Step 2: Convert an image with a pixel size of... 24-bit color host image Layered host image divided into three single color channels: red, green, and blue. Then, each layer of host image of The region is divided into areas of size Non-overlapping image patches, where, For color host image The number of edge pixels, These represent the red, green, and blue channels, respectively. Step 3: In the layered host image Select an image patch sequentially from top to bottom and from left to right. Divide it into sizes of left image patch and size are right-side image patch Based on the derived formulas (1) and (2) for the fusion field of invertible matrices, respectively... and Achieving high-energy focusing yields biaxial high-energy coefficients and ,in, These represent the red, green, and blue channels, respectively. (1) (2) in, Represents image blocks No. Line number Column pixel values, Represents image blocks No. Line number The pixel values ​​of the column; here, let's assume the selected image patch. , , Then we get , ; Step 4: Calculate the biaxial high-energy coefficients according to formulas (3) and (4). and remainder and ; (3) (4) in, It is a modulo function. For the quantization step size of different channels, , This serves as the global baseline step size. These represent the red, green, and blue channels respectively; here, let's set... , Then we get , , ; Step 5: Define the embedded quantization point according to formula (5). ; (5) Step 6: From the layered watermark bit sequence In the process, extract one watermark information to be embedded in sequence. According to the embedded watermark information At the defined embedding quantization point Select the watermark information The 8 unique corresponding quantization points; calculate the Euclidean distance according to formula (6) and return the position index of the quantization point corresponding to the shortest Euclidean distance. ; (6) in, For the quantization step size of different channels, , This serves as the global baseline step size. for coordinates in For the position index of the coordinates, An index function that returns the minimum value. These represent the red, green, and blue channels respectively; here, let's set... , , Then we get ; Step 7: Obtain the coordinates of the nearest quantization point according to formula (7). As a target quantification point; (7) Step 8: According to formulas (8) and (9), respectively... and High energy coefficients with watermarking were obtained by implementing dual-axis collaborative quantization index modulation. and Here, let's assume Then we get ; (8) (9) The meanings of the variables in formulas (8) and (9) are as described above; here, let... , , , , Then we get , ; Step 9: Based on the derived formulas (10) and (11) for allocating pixel changes in the invertible matrix fusion domain, the watermarked floating-point image block is obtained. and ; (10) (11) Here, let , , , Then we get , ; Step 10: Convert the watermarked floating-point image block matrix and Rounding yields an integer image patch matrix containing the watermark. and ,Will and merged into of And update it to its hierarchical host image. The corresponding position in, where, These represent the red, green, and blue channels, respectively. Here, let , Then we get ; Step 11: Repeat steps 3 through 10 of this process until all watermark information has been embedded, thus obtaining a watermarked layered host image. And the layered host image containing the watermark's three red, green, and blue single-color channels Reassemble and obtain pixel size Watermarked images ,in, These represent the red, green, and blue channels, respectively. The watermark extraction process is described as follows: Step 1: Convert an image with a pixel size of 24-bit color watermarked image A layered watermarked image consisting of three single color channels: red, green, and blue. Then, each layer of watermarked image... of The region is divided into areas of size Non-overlapping watermarked image blocks; among which, These represent the red, green, and blue channels, respectively. Step 2: Layering the watermarked image Select a watermarked image block in the order from top to bottom and from left to right. Divide it into sizes of The watermarked image block on the left and size are The watermarked image block on the right Based on the derived formulas (12) and (13) for the fusion field of invertible matrices, respectively... and Achieving high-energy focusing yields biaxial high-energy coefficients and ; (12) (13) in, Represents image blocks No. Line number Column pixel values, Represents watermarked image blocks No. Line number The pixel values ​​of the column; here, let's assume... , , Then we get , ; Step 3: Calculate the biaxial high energy coefficients according to formulas (14) and (15) respectively. and remainder and ; (14) (15) in, It is a modulo function. For the quantization step size of different channels, , This serves as the global baseline step size. These represent the red, green, and blue channels respectively; here, let's set... Then we get , , ; Step 4: Define the extraction quantization points according to formula (16). ; (16) Step 5: Calculate the Euclidean distance according to formula (17) and return the position index of the quantized point corresponding to the shortest Euclidean distance. ; (17) in, For the quantization step size of different channels, , This serves as the global baseline step size. for coordinates For the position index of the coordinates, An index function that returns the minimum value. These represent the red, green, and blue channels respectively; here, let's set... , , , , Then we get ; Step 6: Extract watermark information according to formula (18) ; (18) Here, let Then we get ; Step 7: Repeat steps 2 through 6 of this process to extract the quaternary watermark bit sequence for each layer. Then, each group of 4 quaternary bits is converted into an 8-bit binary sequence, and then into a decimal pixel value. These represent the red, green, and blue channels, respectively. Step 8: Perform a four-key-based process on each layer of decimal pixels after conversion. , , , Decrypting the four-dimensional Lorenz chaotic map and obtaining layered watermarks ,in, These represent the red, green, and blue channels, respectively. Step 9: Combine layered watermarks Forming the final watermark extraction ,in, These represent the red, green, and blue channels, respectively.

[0011] This method employs a dual-axis collaborative mechanism to synchronously modulate the high-energy coefficients of the left and right regions of the image block, and completes quantization index modulation with a linkage structure to achieve watermark information embedding and blind extraction. This method has good watermark invisibility and robustness, and is simple and fast.

Claims

1. The purpose of this invention is to provide a color image blind watermarking method based on dual-axis collaborative quantization index modulation, characterized in that: This is achieved through a specific watermark embedding and extraction process, the watermark embedding process of which is described as follows: Step 1: Convert an image with a pixel size of 24-bit color watermarked image Layered watermark image consisting of three single color channels: red, green, and blue. ; Then, each single color channel is sequentially... Perform based on 4 keys , , , The four-dimensional Lorenz chaotic map is encrypted, and each encrypted decimal pixel value is converted into an 8-bit binary sequence, then these 8-bit binary sequences are converted into a 4-bit quaternary sequence; then, The quaternary sequences in the sequence are concatenated to form a length of . string sequence ,in, For color watermarked images The number of edge pixels, These represent the red, green, and blue channels, respectively. Step 2: Convert an image with a pixel size of... 24-bit color host image Layered host image divided into three single color channels: red, green, and blue. Then, each layer of host image of The region is divided into areas of size Non-overlapping image patches, where, For color host image The number of edge pixels, These represent the red, green, and blue channels, respectively. Step 3: In the layered host image Select an image patch sequentially from top to bottom and from left to right. Divide it into sizes of left image patch and size are right-side image patch Based on the derived formulas (1) and (2) for the fusion field of invertible matrices, respectively... and Achieving high-energy focusing yields biaxial high-energy coefficients and ,in, These represent the red, green, and blue channels, respectively. （1） （2） in, Represents image blocks No. Line number Column pixel values, Represents image blocks No. Line number The pixel values ​​of the column; Step 4: Calculate the biaxial high-energy coefficients according to formulas (3) and (4). and remainder and ; （3） （4） in, It is a modulo function. For the quantization step size of different channels, , This is the global baseline step size. These represent the red, green, and blue channels, respectively. Step 5: Define the embedding quantization point according to formula (5). ; （5） Step 6: From the layered watermark bit sequence In the process, extract one watermark information to be embedded in sequence. According to the embedded watermark information At the defined embedding quantization point Select the watermark information The 8 unique corresponding quantization points; calculate the Euclidean distance according to formula (6) and return the position index of the quantization point corresponding to the shortest Euclidean distance. ; （6） in, For the quantization step size of different channels, , This is the global baseline step size. for coordinates in For the position index of the coordinates, An index function that returns the minimum value. These represent the red, green, and blue channels, respectively. Step 7: Obtain the coordinates of the nearest quantization point according to formula (7). As a target quantification point; （7） Step 8: According to formulas (8) and (9), respectively... and High energy coefficients with watermarking were obtained by implementing dual-axis collaborative quantization index modulation. and ; （8） （9） The meanings of the variables in formulas (8) and (9) are as described above; Step 9: Based on the derived formulas (10) and (11) for allocating pixel changes in the invertible matrix fusion domain, the watermarked floating-point image block is obtained. and ; （10） （11） Step 10: Convert the watermarked floating-point image block matrix and Rounding yields an integer image patch matrix containing the watermark. and ,Will and merged into of And update it to its hierarchical host image. The corresponding position in, where, These represent the red, green, and blue channels, respectively. Step 11: Repeat steps 3 through 10 of this process until all watermark information has been embedded, thus obtaining a watermarked layered host image. And the layered host image containing the watermark's three red, green, and blue single-color channels Reassemble and obtain pixel size Watermarked images ,in, These represent the red, green, and blue channels, respectively. The watermark extraction process is described as follows: Step 1: Convert an image with a pixel size of 24-bit color watermarked image A layered watermarked image consisting of three single color channels: red, green, and blue. Then, each layer of watermarked image... of The region is divided into areas of size Non-overlapping watermarked image blocks; among which, These represent the red, green, and blue channels, respectively. Step 2: Layering the watermarked image Select a watermarked image block in the order from top to bottom and from left to right. Divide it into sizes of The watermarked image block on the left and size are The watermarked image block on the right Based on the derived formulas (12) and (13) for the fusion field of invertible matrices, respectively... and Achieving high-energy focusing yields biaxial high-energy coefficients and ; （12） （13） in, Represents image blocks No. Line number Column pixel values, Represents watermarked image blocks No. Line number The pixel values ​​of the column; Step 3: Calculate the biaxial high energy coefficients according to formulas (14) and (15) respectively. and remainder and ; （14） （15） in, It is a modulo function. For the quantization step size of different channels, , This is the global baseline step size. These represent the red, green, and blue channels, respectively. Step 4: Define the extraction quantization points according to formula (16). ; （16） Step 5: Calculate the Euclidean distance according to formula (17) and return the position index of the quantized point corresponding to the shortest Euclidean distance. ; （17） in, For the quantization step size of different channels, , This is the global baseline step size. for coordinates For the position index of the coordinates, An index function that returns the minimum value. These represent the red, green, and blue channels, respectively. Step 6: Extract watermark information according to formula (18) ; （18） Step 7: Repeat steps 2 through 6 of this process to extract the quaternary watermark bit sequence for each layer. Then, each group of 4 quaternary bits is converted into an 8-bit binary sequence, and then into a decimal pixel value. These represent the red, green, and blue channels, respectively. Step 8: Perform a four-key-based process on each layer of decimal pixels after conversion. , , , Decrypting the four-dimensional Lorenz chaotic map and obtaining layered watermarks ,in, These represent the red, green, and blue channels, respectively. Step 9: Combine layered watermarks Forming the final watermark extraction ,in, These represent the red, green, and blue channels, respectively.

Images