2D digital human rendering watermark encryption method and system

Abstract

This invention relates to the field of digital image processing technology, and particularly to a method and system for encrypting watermarks in 2D digital human rendering. The method includes acquiring encrypted watermark information; acquiring key feature blocks of a digital human rendered image; decomposing the key feature blocks into low-frequency and high-frequency layers using a multi-resolution transformation method; and allocating and embedding the encrypted watermark information into the low-frequency and high-frequency layers of the key feature blocks. This invention combines illumination analysis, complex texture analysis, and dynamic change recognition to effectively extract key feature blocks from digital human rendered images, including facial, limb, and clothing detail areas. This ensures that the watermark embedding can be accurately positioned in important visual areas of the image. By using a multi-resolution transformation method to decompose the key feature blocks into low-frequency and high-frequency layers and rationally allocating the encrypted watermark information to each layer according to sequential intervals, the watermark achieves good stability in the low-frequency layer.

Description

Technical Field

[0001] This invention relates to the field of digital image processing technology, specifically to a 2D digital human rendering watermark encryption method and system. Background Technology

[0002] With the widespread application of 2D digital human rendering technology, digital human images have significant value in film and television production, virtual anchors, digital entertainment, and copyright protection. However, existing digital human rendered images still have obvious technical shortcomings in terms of copyright protection and watermark embedding. Traditional digital image watermarking methods are mostly aimed at static images, usually embedding the watermark into the background or a local area of ​​the entire image, without fully considering the complexity and dynamism of key feature blocks in digital human images. For example, details such as facial expressions, body movements, and clothing wrinkles are extremely important in image representation. Existing technologies ignore the dynamic changes of human features during the rendering process, resulting in insufficient accuracy in extracting key feature blocks, and watermarks are easily tampered with or lost after image compression and cropping.

[0003] Furthermore, existing watermark embedding methods typically employ a single-resolution embedding strategy, which involves directly embedding watermark information into image pixels or grayscale coefficients. This results in a lack of hierarchy in the watermark, making it difficult to fully utilize low-frequency structural information. For dynamically rendered digital human images, existing technologies cannot effectively capture changes in facial expressions, arm movements, or other dynamic feature blocks, making it difficult for the watermark to be evenly distributed in key feature areas and reducing the watermark's concealment and anti-tampering capabilities. Summary of the Invention

[0004] To address the aforementioned problems, this invention provides a 2D digital human rendering watermark encryption method and system.

[0005] This invention employs the following technical solution: a 2D digital human rendering watermark encryption method, comprising:

[0006] The copyright identifier, timestamp, and related metadata are converted into encrypted watermark information using encryption algorithms.

[0007] Obtain key feature blocks from the rendered image of the digital human;

[0008] By combining multi-resolution transformation methods of images, key feature blocks are decomposed into low-frequency and high-frequency layers; encrypted watermark information is allocated and embedded in the low-frequency and high-frequency layers of key feature blocks.

[0009] After the encrypted watermark information is successfully embedded in multiple layers of the image, the key feature blocks after watermark embedding are merged with the original image through an inverse transformation process to reconstruct a complete 2D digital human rendering image.

[0010] The method for obtaining key feature blocks of a digital human rendered image includes:

[0011] The digital human rendered image is preprocessed using Gaussian filtering, mean filtering, or median filtering to remove noise from the image.

[0012] Identify areas of lighting variation in rendered images of digital humans;

[0013] Identify complex texture regions in rendered images of digital humans;

[0014] Mark all pixels in the intersection of the region with the lighting change area and the region with the complex texture, and extract the lighting coefficients and complex texture coefficients of all pixels;

[0015] The illumination coefficient and the complex texture coefficient are weighted and fused to obtain the comprehensive feature value of each pixel;

[0016] Set a feature threshold, mark pixels with a comprehensive feature value greater than the feature threshold as feature pixels, extract feature pixels, and map the position of feature pixels in the digital human rendering image to a blank background layer to obtain multiple extraction patterns with intervals composed of feature pixels.

[0017] Obtain the number of pixels in multiple extracted patterns, preset a pixel count gradient threshold, and mark the extracted patterns that fall within the pixel count gradient threshold as key feature blocks.

[0018] As a further description of the above technical solution: the relevant metadata includes creator, copyright information, creation time and usage license data.

[0019] As a further description of the above technical solution: the method for converting into encrypted watermark information includes:

[0020] Obtain copyright information, timestamps, and related metadata associated with the digital human rendered image, mark them as watermark data, and convert the watermark data into XML format via user input.

[0021] The acquired watermark data is constructed into a complete watermark data set, and the watermark data is preprocessed.

[0022] A symmetric encryption algorithm is used for encryption. First, an encryption key is generated, and then the watermark data is encrypted using the corresponding encryption key to generate encrypted watermark information.

[0023] As a further description of the above technical solution: the method for obtaining the illumination change area includes:

[0024] Obtain the horizontal and vertical gradients of each pixel in the rendered image of the digital human;

[0025] The illumination coefficient is calculated based on the acquired horizontal and vertical gradients.

[0026] A preset illumination coefficient threshold is used to mark all pixels in the image whose illumination coefficient is greater than the threshold as areas of illumination variation.

[0027] As a further description of the above technical solution: the method for obtaining the complex texture region includes:

[0028] The gray-level relationship between each pixel in the digital human rendered image and its corresponding neighboring pixels is calculated to generate a gray-level co-occurrence matrix;

[0029] Texture features, including contrast and entropy, are extracted based on the gray-level co-occurrence matrix.

[0030] For each pixel in the image, select the corresponding neighboring pixels and calculate the binary relationship between the gray values ​​of the pixel and the neighboring pixels;

[0031] Based on the binary relationship, a binary code is generated, and the binary code is converted into a decimal number. The decimal number is then used as the LBP value of the corresponding pixel.

[0032] The contrast, entropy and LBP values ​​of each pixel are weighted and fused to obtain the complex texture coefficients of each pixel.

[0033] A preset texture threshold is used to mark pixels with a complex texture coefficient greater than the texture threshold as complex texture regions.

[0034] As a further description of the above technical solution: the method for obtaining key feature blocks of a digital human rendered image also includes:

[0035] The 2D digital human rendered image sequence is preprocessed using methods such as Gaussian filtering, mean filtering, or median filtering to remove image noise and obtain the previous and next frame images of the current frame image.

[0036] Obtain the pixel points of key points in the digital human rendering image; the key points include the face, hands, feet, and torso;

[0037] Motion analysis of pixels at each key point in the digital human rendered image is performed over multiple consecutive frames to calculate the change range of pixel grayscale and obtain the pixel-level dynamic change coefficient.

[0038] A preset dynamic change coefficient threshold is used to mark pixels with a dynamic change coefficient greater than the threshold as dynamically changing pixels.

[0039] The positions of dynamically changing pixels in the digital human rendered image are mapped onto a blank background layer to obtain multiple extraction patterns with intervals composed of dynamically changing pixels.

[0040] Obtain the number of pixels in multiple extracted patterns, preset a pixel count gradient threshold, and mark the extracted patterns that fall within the pixel count gradient threshold as key feature blocks.

[0041] As a further description of the above technical solution: the method of allocating encrypted watermark information into the low-frequency layer and high-frequency layer of key feature blocks includes:

[0042] For the key feature blocks obtained, DWT wavelet decomposition is used to decompose the image into low-frequency and high-frequency layers;

[0043] The encrypted watermark information is encoded into a binary bit sequence, numbered sequentially, with each bit corresponding to an index. Even-numbered indices are embedded in the low-frequency layer, and odd-numbered indices are embedded in the high-frequency layer.

[0044] After the encrypted watermark information is successfully embedded in multiple layers of the image, the watermarked image is restored to a complete digital human rendering image through an inverse transformation process.

[0045] As a further description of the above technical solution: the method of using DWT wavelet decomposition to decompose an image into a low-frequency layer and a high-frequency layer includes:

[0046] Haar wavelets were chosen as the basis functions for DWT.

[0047] Two-dimensional discrete wavelet transform is performed on the image region of each key feature block, including low-pass filtering and high-pass filtering of the image rows to generate two intermediate matrices, and column-direction low-pass filtering and high-pass filtering of the two intermediate matrices to obtain four sub-bands;

[0048] The four subbands are: LL subband: contains overall image structure information; LH subband: contains horizontal texture information; HL subband: contains vertical texture information; HH subband: contains diagonal details and texture information;

[0049] The LL subband is used as the low-frequency layer, and the LH, HL, and HH subbands are used as the high-frequency layers.

[0050] A 2D digital human rendering watermarking encryption system is provided for implementing the aforementioned 2D digital human rendering watermarking encryption method. The system includes:

[0051] The encrypted watermark information acquisition module uses an encryption algorithm to convert copyright identifiers, timestamps, and related metadata into encrypted watermark information.

[0052] The key feature block acquisition module acquires key feature blocks from the rendered image of the digital human.

[0053] The decomposition and embedding module, combined with the multi-resolution transformation method of the image, decomposes the key feature blocks into low-frequency and high-frequency layers, and distributes the encrypted watermark information into the low-frequency and high-frequency layers of the key feature blocks.

[0054] The inverse transform merging module, after successfully embedding the encrypted watermark information into multiple layers of the image, merges the key feature blocks embedded with the watermark with the original image through an inverse transform process to reconstruct a complete 2D digital human rendering image.

[0055] The beneficial effects of this invention are as follows:

[0056] This invention effectively extracts key feature blocks from digital human rendered images by combining illumination analysis, complex texture analysis, and dynamic change recognition. These feature blocks include facial, limb, and clothing details, ensuring that the watermark embedding is accurately positioned in important visual areas of the image. A multi-resolution transformation method is used to decompose the key feature blocks into low-frequency and high-frequency layers, and the encrypted watermark information is rationally distributed to each layer according to sequential intervals. This ensures good stability of the watermark in the low-frequency layer and high concealment in the high-frequency layer, allowing the embedded watermark to be reliably extracted even after image compression, cropping, or local modification. Through comprehensive weighting of illumination and texture, the recognition of key feature blocks is more accurate, avoiding errors caused by single facial, hand, and foot feature extraction in existing technologies, and fully preserving image details and visual effects. Attached Figure Description

[0057] The present invention will be further explained below with reference to the accompanying drawings and embodiments:

[0058] Figure 1 This is a flowchart of a 2D digital human rendering watermark encryption method provided in Embodiment 1 of the present invention;

[0059] Figure 2 This is a flowchart of obtaining key feature blocks of a digital human rendered image provided in Embodiment 1 of the present invention;

[0060] Figure 3 This is a flowchart of obtaining key feature blocks of a digital human rendered image provided in Embodiment 2 of the present invention;

[0061] Figure 4 This is a module connection diagram of a 2D digital human rendering watermark encryption system provided in Embodiment 3 of the present invention. Detailed Implementation

[0062] To make the technical means, creative features, objectives, and effects of this invention readily understandable, the invention is further described below with reference to specific illustrations. It should be noted that, unless otherwise specified, the embodiments and features described in these embodiments can be combined with each other.

[0063] Example 1

[0064] Please see Figures 1-2 This invention provides a technical solution: a 2D digital human rendering watermark encryption method, comprising:

[0065] The copyright identifier, timestamp, and related metadata are converted into encrypted watermark information using encryption algorithms.

[0066] The relevant metadata includes creator information, copyright information, creation time, and usage license data.

[0067] The method for converting the information into encrypted watermark information includes:

[0068] Obtain copyright information, timestamps, and related metadata associated with digital images, mark them as watermark data, and convert the watermark data into a standard format (e.g., JSON or XML) through user input.

[0069] The acquired watermark data is constructed into a complete watermark data set, and the watermark data is preprocessed to meet the input requirements of the encryption algorithm.

[0070] The preprocessing includes data encoding (such as UTF-8 encoding) and removing extra spaces and special characters.

[0071] The watermark data is encrypted using a symmetric encryption algorithm (such as AES). First, an encryption key is generated, and then the watermark data is encrypted using the corresponding encryption key to generate encrypted watermark information.

[0072] It should be noted that the use of symmetric encryption algorithms (such as AES) for encryption is an existing technology and will not be elaborated upon here.

[0073] Obtain key feature blocks from the rendered image of the digital human;

[0074] It should be noted that digital human images typically contain details such as faces and limbs, and these feature blocks play a crucial role in the image's representation. To protect the content within these key feature blocks, especially when embedding watermarks or protecting digital rights, accurately acquiring these feature blocks is one of the key technologies. In existing technologies, most methods for acquiring feature blocks in images rely on static image processing algorithms, which fail to fully consider the changes in human features during dynamic rendering, leading to insufficient accuracy in feature block extraction.

[0075] Methods for obtaining key feature blocks in digital human rendered images include:

[0076] The digital human rendered image is preprocessed using Gaussian filtering, mean filtering, or median filtering to remove noise from the image.

[0077] Identify areas of lighting variation in rendered images of digital humans;

[0078] The method for obtaining the illumination variation region includes:

[0079] Obtain the horizontal and vertical gradients of each pixel in the rendered image of the digital human;

[0080] The formula for calculating the horizontal gradient is: In the formula, The gradient is in the horizontal direction. It is the grayscale value of the pixel (x,y) in the digital human rendered image;

[0081] The formula for calculating the vertical gradient is: In the formula, The gradient is in the vertical direction. It is the grayscale value of the pixel (x,y) in the digital human rendered image;

[0082] The illumination coefficient is calculated based on the acquired horizontal and vertical gradients.

[0083] The formula for calculating the illumination coefficient is: In the formula, The illumination coefficient is the value at pixel (x, y).

[0084] A preset illumination coefficient threshold is used to mark all pixels in the image whose illumination coefficient is greater than the threshold as areas of illumination variation.

[0085] Identify complex texture regions in rendered images of digital humans;

[0086] The method for obtaining the complex texture region includes:

[0087] The gray-level relationship between each pixel in the digital human rendered image and its corresponding neighboring pixels is calculated to generate a gray-level co-occurrence matrix;

[0088] It should be noted that the gray-level co-occurrence matrix can describe the gray-level changes between pixel pairs, reflecting the repeatability and directionality of the texture. In this embodiment, neighboring pixels are selected in four directions: 0°, 45°, 90° and 135° to calculate the gray-level co-occurrence matrix.

[0089] Example: Assume the image grayscale range is (0, G−1), the neighborhood direction can be 0°, 45°, 90°, 135°, and the distance is 1. Select the current pixel (x,y) and its right neighboring pixel with a direction of 0° and a distance of 1.

[0090] If the gray level of the current pixel is i and the gray level of the neighboring pixel to the right is j, then the matrix element P[i,j]+1;

[0091] After statistical analysis of the entire image, normalization is performed: The gray-level co-occurrence matrix is ​​obtained as follows: ;

[0092] Texture features, including contrast and entropy, are extracted based on the gray-level co-occurrence matrix.

[0093] It should be noted that contrast is used to measure the intensity of pixel grayscale changes, while entropy measures the irregularity of texture; the higher the entropy, the more complex the texture.

[0094] The formula for calculating contrast is: ;

[0095] The formula for calculating entropy is: ;

[0096] In the formula, This is used to avoid taking the logarithm for zero probability.

[0097] For each pixel in the image, select the corresponding neighboring pixels (usually a 3×3 neighborhood) and calculate the binary relationship between the pixel and the grayscale values ​​of the neighboring pixels;

[0098] Example: Assume the grayscale of a 3×3 neighborhood pixel is: ;

[0099] The center pixel is set to Pc=65;

[0100] The corresponding neighboring pixel gray levels are numbered clockwise as follows:

[0101] P0=50, P1=80, P2=70, P3=75, P4=72, P5=68, P6=55, P7=60;

[0102] Compare the grayscale of the center pixel with the corresponding grayscale of the neighboring pixels one by one. If the grayscale of the center pixel is greater than or equal to that of the neighboring pixels, the corresponding position of the binary relationship is set to 0. If the grayscale of the center pixel is less than that of the neighboring pixels, the corresponding position of the binary relationship is set to 1. The resulting binary code is: 01111100.

[0103] Based on the binary relationship, a binary code is generated, and the binary code is converted into a decimal number. The decimal number is then used as the LBP value of the corresponding pixel.

[0104] The contrast, entropy and LBP values ​​of each pixel are weighted and fused to obtain the complex texture coefficients of each pixel.

[0105] Optionally, the method for calculating the complex texture coefficients is as follows:

[0106] ;

[0107] In the formula, For pixels Complex texture coefficients, For pixels Contrast For pixels entropy, For pixels LBP value, , and The weighting coefficients are used for calculation. It should be noted that the formulas mentioned above are all numerical calculations after removing dimensions. They are formulas that are closest to the real situation, obtained by software simulation based on a large amount of data. The weighting coefficients and the preset thresholds in the analysis process are set by those skilled in the art based on the actual situation or obtained by simulation based on a large amount of data. The size of the weighting coefficient is a specific value obtained by quantifying each parameter to facilitate subsequent comparison. The size of the weighting coefficient depends on the amount of sample data and the processing coefficients initially set by those skilled in the art for each set of sample data. As long as it does not affect the proportional relationship between the parameter and the quantified value, it is acceptable.

[0108] A preset texture threshold is used to mark pixels with a complex texture coefficient greater than the texture threshold as complex texture regions.

[0109] Mark all pixels in the intersection of the region with the lighting change area and the region with the complex texture, and extract the lighting coefficients and complex texture coefficients of all pixels;

[0110] The illumination coefficient and the complex texture coefficient are weighted and fused to obtain the comprehensive feature value of each pixel;

[0111] Example: The formula for calculating the comprehensive feature value is:

[0112] ;

[0113] In the formula, For pixels The comprehensive eigenvalues, For pixels Illuminance coefficient, For pixels Complex texture coefficients, and These are the weighting coefficients.

[0114] Set a feature threshold, mark pixels with a comprehensive feature value greater than the feature threshold as feature pixels, extract the feature pixels, and map the position of the feature pixels in the digital human rendering image onto a blank background layer to obtain multiple extraction patterns with intervals composed of feature pixels.

[0115] Obtain the number of pixels in multiple extracted patterns, preset a pixel count gradient threshold, and mark the extracted patterns that fall within the pixel count gradient threshold as key feature blocks;

[0116] It should be noted that the aforementioned thresholds were set by those skilled in the art based on actual circumstances or obtained through simulation with a large amount of data.

[0117] In this embodiment, by combining illumination analysis and texture analysis, regions with significant illumination changes and complex textures in an image can be effectively identified. These regions are of great significance for watermark embedding and can provide higher concealment. By fusing illumination and texture features, key detail regions in the image can be accurately located, providing effective support for copyright protection, authentication, and content recognition of digital images.

[0118] By combining multi-resolution transformation methods of images, key feature blocks are decomposed into low-frequency and high-frequency layers; encrypted watermark information is allocated and embedded in the low-frequency and high-frequency layers of key feature blocks.

[0119] Specifically, the method for allocating encrypted watermark information into low-frequency and high-frequency layers embedded in key feature blocks includes:

[0120] For the key feature blocks obtained, DWT wavelet decomposition is used to decompose the image into a low-frequency (LL) layer and a high-frequency (LH, HL, HH) layer. The low-frequency layer contains the overall structural information of the image, while the high-frequency layer contains details and texture information.

[0121] The encrypted watermark information is encoded into a binary bit sequence, numbered sequentially, with each bit corresponding to an index. Even-numbered indices are embedded in the low-frequency layer, and odd-numbered indices are embedded in the high-frequency layer.

[0122] After the encrypted watermark information is successfully embedded in multiple layers of the image, the key feature blocks after watermark embedding are merged with the original image through an inverse transformation process to reconstruct a complete 2D digital human rendering image.

[0123] Specifically, it should be noted that after embedding the low-frequency and high-frequency watermarks, the multi-resolution transformed images need to be recombined to restore the complete 2D digital human rendering image. This process is usually achieved through inverse multi-resolution transformation, which synthesizes the data from the low-frequency and high-frequency layers and restores them to the spatial domain of the original image. Inverse transformation is an existing technology and will not be specifically described in this invention. The implementation method can use conventional inverse wavelet or inverse pyramid algorithms, the functions of which are: to maintain the integrity of the key feature blocks of the embedded watermark; to restore the visual effect of the image so that the image after watermark embedding is visually indistinguishable from the original image; and to ensure that the multi-level embedded watermark data can still be extracted in the spatial domain.

[0124] For the acquired key feature blocks, the method of decomposing the image into low-frequency (LL) and high-frequency (LH, HL, HH) layers using DWT wavelet decomposition includes:

[0125] Choosing the Haar wavelet as the basis function for DWT facilitates fast computation while preserving image edge and texture information;

[0126] Obtain the pixel coordinates and area of ​​key feature blocks;

[0127] Two-dimensional discrete wavelet transform is performed on the image region of each key feature block, including low-pass filtering and high-pass filtering of the image rows to generate two intermediate matrices, and column-direction low-pass filtering and high-pass filtering of the two intermediate matrices to obtain four sub-bands;

[0128] The four subbands are: LL subband: contains overall image structure information; LH subband: contains horizontal texture information; HL subband: contains vertical texture information; HH subband: contains diagonal details and texture information;

[0129] The LL subband is used as the low-frequency layer, and the LH, HL, and HH subbands are used as the high-frequency layers.

[0130] Example 2

[0131] Please see Figure 3 The present invention provides a technical solution: the method for obtaining key feature blocks of a digital human rendered image further includes:

[0132] The 2D digital human rendered image sequence is preprocessed using methods such as Gaussian filtering, mean filtering, or median filtering to remove image noise and obtain the previous and next frame images of the current frame image.

[0133] Obtain the pixel points of key points in the digital human rendering image; the key points include the face, hands, feet, and torso;

[0134] Motion analysis of pixels at each key point in the digital human rendered image is performed over multiple consecutive frames to calculate the change range of pixel grayscale and obtain the pixel-level dynamic change coefficient.

[0135] The expression for the dynamic change coefficient is: In the formula, For pixels The dynamic change coefficient, For the current frame pixel pixel grayscale values, For the previous frame's pixels The pixel grayscale value, where N is the frame number.

[0136] A preset dynamic change coefficient threshold is used to mark pixels with a dynamic change coefficient greater than the threshold as dynamically changing pixels.

[0137] The positions of dynamically changing pixels in the digital human rendered image are mapped onto a blank background layer to obtain multiple extraction patterns with intervals composed of dynamically changing pixels.

[0138] Obtain the number of pixels in multiple extracted patterns, preset a pixel count gradient threshold, and mark the extracted patterns that fall within the pixel count gradient threshold as key feature blocks.

[0139] It should be noted that the aforementioned thresholds were set by those skilled in the art based on actual circumstances or obtained through simulation with a large amount of data.

[0140] In this embodiment, key feature blocks of the digital human rendered image are obtained, which can accurately extract regions with significant dynamic changes in the image, including key action parts such as facial expressions, arms, feet, and torso. Motion analysis is performed using a series of consecutive frames, and pixel-level dynamic change coefficients are calculated. This effectively distinguishes between regions with large movements and static or slightly moving regions, thus accurately marking dynamically changing pixels at the pixel level. By mapping these marked dynamically changing pixels to a blank background layer and performing region aggregation and filtering, multiple feature extraction patterns with intervals can be formed, further obtaining precise key feature blocks. The key feature blocks extracted using this method can completely preserve the motion information of the digital human during the rendering process, providing stable and high-value target regions for watermark embedding. Compared with existing technologies, this method fully considers the continuous changes in human features during dynamic rendering, improving the accuracy and robustness of key feature blocks, thereby enhancing the concealment and anti-tampering capabilities of watermark embedding. This provides a reliable and effective technical means for copyright protection, identity verification, and content recognition of digital human rendered images.

[0141] Example 3

[0142] Please see Figure 4 This invention provides a technical solution: a 2D digital human rendering watermark encryption system, which is used to implement the aforementioned 2D digital human rendering watermark encryption method. The system includes:

[0143] The encrypted watermark information acquisition module uses an encryption algorithm to convert copyright identifiers, timestamps, and related metadata into encrypted watermark information.

[0144] The key feature block acquisition module acquires key feature blocks from the rendered image of the digital human.

[0145] The decomposition and embedding module, combined with the multi-resolution transformation method of the image, decomposes the key feature blocks into low-frequency and high-frequency layers, and distributes the encrypted watermark information into the low-frequency and high-frequency layers of the key feature blocks.

[0146] The inverse transform merging module, after successfully embedding the encrypted watermark information into multiple layers of the image, merges the key feature blocks embedded with the watermark with the original image through an inverse transform process to reconstruct a complete 2D digital human rendering image.

[0147] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of the present invention is defined by the appended claims and their equivalents.

Claims

1. A 2D digital human rendering watermark encryption method, characterized in that, include: The copyright identifier, timestamp, and related metadata are converted into encrypted watermark information using encryption algorithms. Obtain key feature blocks from the rendered image of the digital human; By combining multi-resolution transformation methods of images, key feature blocks are decomposed into low-frequency and high-frequency layers; encrypted watermark information is allocated and embedded into the low-frequency and high-frequency layers of key feature blocks; After the encrypted watermark information is successfully embedded in multiple layers of the image, the key feature blocks after watermark embedding are merged with the original image through an inverse transformation process to reconstruct a complete 2D digital human rendering image. The method for obtaining key feature blocks of a digital human rendered image includes: Preprocess the rendered images of the digital human; Identify areas of varying lighting and complex textures in rendered images of digital humans; Mark all pixels in the intersection of the lighting variation area and the complex texture area, and extract the lighting coefficients and complex texture coefficients of all pixels; The illumination coefficient and complex texture coefficient are weighted and fused to obtain the comprehensive feature value of each pixel. A feature threshold is set, and pixels with comprehensive feature values ​​greater than the feature threshold are marked as feature pixels. The feature pixels are extracted, and the positions of the feature pixels in the digital human rendering image are mapped to a blank background layer to obtain multiple extraction patterns with intervals composed of feature pixels. Obtain the number of pixels in multiple extracted patterns, preset a pixel count gradient threshold, and mark the extracted patterns that fall within the pixel count gradient threshold as key feature blocks.

2. The 2D digital human rendering watermark encryption method according to claim 1, characterized in that, The relevant metadata includes creator information, copyright information, creation time, and usage license data.

3. The 2D digital human rendering watermark encryption method according to claim 1, characterized in that, The method for converting the information into encrypted watermark information includes: Obtain copyright information, timestamps, and related metadata associated with the digital human rendered image, mark them as watermark data, and convert the watermark data into XML format via user input. The acquired watermark data is constructed into a complete watermark data set, and the watermark data is preprocessed. A symmetric encryption algorithm is used for encryption. First, an encryption key is generated, and then the watermark data is encrypted using the corresponding encryption key to generate encrypted watermark information.

4. The 2D digital human rendering watermark encryption method according to claim 1, characterized in that, The method for obtaining the illumination variation region includes: Obtain the horizontal and vertical gradients of each pixel in the rendered image of the digital human; The illumination coefficient is calculated based on the acquired horizontal and vertical gradients. A preset illumination coefficient threshold is used to mark all pixels in the image whose illumination coefficient is greater than the threshold as areas of illumination variation.

5. The 2D digital human rendering watermark encryption method according to claim 1, characterized in that, The method for obtaining the complex texture region includes: The gray-level relationship between each pixel in the digital human rendered image and its corresponding neighboring pixels is calculated to generate a gray-level co-occurrence matrix; Texture features, including contrast and entropy, are extracted based on the gray-level co-occurrence matrix. For each pixel in the image, select the corresponding neighboring pixels and calculate the binary relationship between the gray values ​​of the pixel and its neighboring pixels; Based on the binary relationship, a binary code is generated, and the binary code is converted into a decimal number. The decimal number is then used as the LBP value of the corresponding pixel. The contrast, entropy and LBP values ​​of each pixel are weighted and fused to obtain the complex texture coefficients of each pixel. A preset texture threshold is used to mark pixels with a complex texture coefficient greater than the texture threshold as complex texture regions.

6. The 2D digital human rendering watermark encryption method according to claim 1, characterized in that, The method for obtaining key feature blocks of a digital human rendered image also includes: The 2D digital human rendering image sequence is preprocessed to obtain the pixel points of key points in the digital human rendering image, including the face, hands, feet and torso; Motion analysis is performed on the pixels of each key point for multiple consecutive frames to obtain the pixel-level dynamic change coefficient. A dynamic change coefficient threshold is preset, and pixels with dynamic change coefficients greater than the dynamic change coefficient threshold are marked as dynamically changing pixels. The positions of dynamically changing pixels in the digital human rendered image are mapped onto a blank background layer to obtain multiple extraction patterns with intervals composed of dynamically changing pixels. Obtain the number of pixels in multiple extracted patterns, preset a pixel count gradient threshold, and mark the extracted patterns that fall within the pixel count gradient threshold as key feature blocks.

7. The 2D digital human rendering watermark encryption method according to claim 1, characterized in that, The method for allocating and embedding encrypted watermark information into low-frequency and high-frequency layers of key feature blocks includes: For the key feature blocks obtained, DWT wavelet decomposition is used to decompose the image into low-frequency and high-frequency layers; The encrypted watermark information is encoded into a binary bit sequence, numbered sequentially, with each bit corresponding to an index. Even-numbered indices are embedded in the low-frequency layer, and odd-numbered indices are embedded in the high-frequency layer. After the encrypted watermark information is successfully embedded in multiple layers of the image, the watermarked image is restored to a complete digital human rendering image through an inverse transformation process.

8. A 2D digital human rendering watermark encryption method according to claim 7, characterized in that, The method of decomposing an image into low-frequency and high-frequency layers using DWT wavelet decomposition includes: Haar wavelets were chosen as the basis functions for DWT. Two-dimensional discrete wavelet transform is performed on the image region of each key feature block, including low-pass filtering and high-pass filtering of the image rows to generate two intermediate matrices, and column-direction low-pass filtering and high-pass filtering of the two intermediate matrices to obtain four sub-bands; The four subbands are: LL subband: contains overall image structure information; LH subband: contains horizontal texture information; HL subband: contains vertical texture information; HH subband: contains diagonal details and texture information; The LL subband is used as the low-frequency layer, and the LH, HL, and HH subbands are used as the high-frequency layers.

9. A 2D digital human rendering watermarking encryption system, used to implement the 2D digital human rendering watermarking encryption method according to any one of claims 1-8, characterized in that, The system includes: The encrypted watermark information acquisition module uses an encryption algorithm to convert copyright identifiers, timestamps, and related metadata into encrypted watermark information. The key feature block acquisition module acquires key feature blocks from the rendered image of the digital human. The decomposition and embedding module, combined with the multi-resolution transformation method of the image, decomposes the key feature blocks into low-frequency and high-frequency layers, and distributes the encrypted watermark information into the low-frequency and high-frequency layers of the key feature blocks. The inverse transform merging module, after successfully embedding the encrypted watermark information into multiple layers of the image, merges the key feature blocks embedded with the watermark with the original image through an inverse transform process to reconstruct a complete 2D digital human rendering image.

Images