Screen hidden watermark extraction tracing method and system, storage medium and electronic device

Abstract

This application provides a method, system, storage medium, and electronic device for extracting and tracing hidden watermarks from a screen, relating to the field of image processing technology. The technical solution provided in this application divides a photographed image into multiple equally sized grid units based on screen parameters, transforming the originally inconsistent overall image quality into independently evaluable local regions. Target grid units with well-preserved watermark information are selected through watermark recognition confidence screening, avoiding the negative impact of low-quality regions on the overall analysis results. Local watermark information is fused based on the splicing positions of multiple adjacent target grid units, compensating for the incompleteness of information in a single region through regional collaboration, thereby improving the completeness and accuracy of the global watermark information. Through a progressive processing approach involving regionalization, quality screening, collaborative fusion, and device optimization, the accuracy of watermark recognition is significantly improved, thereby enhancing the reliability of the tracing analysis.

Description

Technical Field

[0001] This application relates to the field of image processing technology, specifically to a method, system, storage medium, and electronic device for extracting and tracing hidden watermarks on a screen. Background Technology

[0002] Screen watermarking is a technique that uses embedded watermark information to trace the source and dissemination path of images. This technology has significant applications in information security, copyright protection, and leak tracing. Screen watermarks are typically embedded into the displayed image using digital watermarking algorithms during the content display phase. When a user takes a photo or screenshot of the screen, the watermark information is recorded along with the image. By extracting and analyzing the watermark information from leaked images, key tracing information such as the image's shooting device, shooting time, and viewing user can be determined, thereby enabling precise location of the source of the information leak and accountability.

[0003] Current source tracing analysis methods primarily involve uniformly processing the entire image, employing a single watermark extraction algorithm and fixed processing parameters to perform watermark recognition. However, in real-world applications, different areas of a photographed image are significantly affected. Perspective distortion caused by the shooting angle is more severe at the image edges, uneven lighting leads to varying brightness differences between the center and periphery of the image, and localized reflections or smudges on the screen severely damage the watermark information in specific areas. These factors result in inconsistent watermark information quality across different regions of the image. Because existing technologies use a uniform processing strategy for the entire image, they cannot adaptively adjust to the differences in watermark quality across different regions, leading to reduced watermark recognition accuracy and poor reliability of source tracing analysis. Summary of the Invention

[0004] This application provides a method, system, storage medium, and electronic device for extracting and tracing hidden watermarks on a screen, which can improve the accuracy of watermark recognition and thus improve the reliability of tracing analysis.

[0005] Firstly, this application provides a method for extracting and tracing the source of a hidden watermark on a screen, the method comprising: To obtain a reproduced image of the screen after it has been photographed using a camera; Obtain the screen parameters from the reproduced image, and divide the reproduced image into multiple grid units of equal size according to the screen parameters; Perform feature analysis on the image of each grid cell and output the watermark recognition confidence score for each grid cell; Grid cells with a watermark recognition confidence level greater than a preset confidence threshold are identified as target grid cells, and the feature patterns of the target grid cells are matched with preset watermark templates to determine the local watermark information of the target grid cells. Based on the splicing position obtained after merging and splicing multiple adjacent target grid cells, multiple local watermark information is fused to obtain global watermark information; The camera imaging parameters of the shooting device are obtained, and the global watermark information is optimized based on the camera imaging parameters to obtain the target watermark information of the reproduced image. The target watermark information is used to trace the reproduction information of the reproduced image.

[0006] By employing the aforementioned technical solution, the reproduced image is segmented into multiple equally sized grid units based on screen parameters, transforming the originally inconsistent overall image into independently evaluable local regions. Target grid units with well-preserved watermark information are selected through watermark recognition confidence screening, avoiding the negative impact of low-quality regions on the overall analysis results. Local watermark information is fused based on the splicing positions of multiple adjacent target grid units, compensating for the incompleteness of information in a single region through regional collaboration, thereby improving the completeness and accuracy of the global watermark information. Through a progressive processing approach involving regionalization, quality screening, collaborative fusion, and equipment optimization, the accuracy of watermark recognition is significantly improved, thereby enhancing the reliability of source tracing analysis.

[0007] Secondly, this application provides a system for extracting and tracing hidden watermarks on a screen, the system comprising: The image input module is used to acquire a copy of the image generated after the screen is captured by a camera. The grid division module is used to obtain screen parameters from the reproduced image and divide the reproduced image into multiple grid units of equal size according to the screen parameters; The confidence calculation module is used to perform feature analysis on the image of each grid cell and output the watermark recognition confidence score corresponding to each grid cell. The grid filtering module is used to identify grid cells with a watermark recognition confidence level greater than a preset confidence threshold as target grid cells, and to match the feature patterns of the target grid cells with a preset watermark template to determine the local watermark information of the target grid cells. The fusion module is used to fuse multiple local watermark information based on the splicing position obtained after merging and splicing multiple adjacent target grid cells to obtain global watermark information. The optimization processing module is used to obtain the camera imaging parameters of the shooting device, and optimize the global watermark information according to the camera imaging parameters to obtain the target watermark information of the reproduced image. The target watermark information is used to trace the reproduction information of the reproduced image.

[0008] Thirdly, this application provides a computer storage medium that stores multiple instructions adapted for loading by a processor and executing any of the methods described above.

[0009] Fourthly, this application provides an electronic device including a processor, a memory, and a transceiver. The memory is used to store instructions, the transceiver is used to communicate with other devices, and the processor is used to execute the instructions stored in the memory to cause the electronic device to perform any of the methods described above.

[0010] In summary, the beneficial effects of the technical solution of this application include: By employing the aforementioned technical solution, the reproduced image is segmented into multiple equally sized grid units based on screen parameters, transforming the originally inconsistent overall image into independently evaluable local regions. Target grid units with well-preserved watermark information are selected through watermark recognition confidence screening, avoiding the negative impact of low-quality regions on the overall analysis results. Local watermark information is fused based on the splicing positions of multiple adjacent target grid units, compensating for the incompleteness of information in a single region through regional collaboration, thereby improving the completeness and accuracy of the global watermark information. Through a progressive processing approach involving regionalization, quality screening, collaborative fusion, and equipment optimization, the accuracy of watermark recognition is significantly improved, thereby enhancing the reliability of source tracing analysis. Attached Figure Description

[0011] Figure 1 This is a flowchart illustrating a method for extracting and tracing hidden watermarks from a screen according to an embodiment of this application. Figure 2 This is a schematic diagram of the structure of a screen hidden watermark extraction and tracing system according to an embodiment of this application; Figure 3 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application.

[0012] Explanation of reference numerals in the attached drawings: 300, electronic device; 301, processor; 302, communication bus; 303, user interface; 304, network interface; 305, memory. Detailed Implementation

[0013] To enable those skilled in the art to better understand the technical solutions in this specification, the technical solutions in the embodiments of this specification will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments.

[0014] In the description of the embodiments of this application, words such as "illustrative," "for example," or "for example" are used to indicate examples, illustrations, or explanations. Any embodiment or design described as "illustrative," "for example," or "for example" in the embodiments of this application should not be construed as being more preferred or advantageous than other embodiments or design solutions. Rather, the use of words such as "illustrative," "for example," or "for example" is intended to present the relevant concepts in a specific manner.

[0015] In the description of the embodiments of this application, the term "multiple" means two or more. For example, multiple systems means two or more systems, and multiple screen terminals means two or more screen terminals. Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the indicated technical features. Thus, a feature defined with "first" or "second" may explicitly or implicitly include one or more of that feature. The terms "comprising," "including," "having," and variations thereof all mean "including but not limited to," unless otherwise specifically emphasized.

[0016] Please see Figure 1 This is a flowchart illustrating a method for extracting and tracing hidden watermarks on a screen, provided in an embodiment of this application. This method can be implemented using a computer program, a microcontroller, or run on a screen watermark extraction and tracing system based on the von Neumann architecture. The computer program can be integrated into an application or run as a standalone utility application. The specific steps of the screen watermark extraction and tracing method are described in detail below.

[0017] S101: Acquire a reproduced image generated after taking a picture of the screen using a camera; Among them, a photographic image refers to a digital image file formed by taking a picture of the content displayed on a monitor screen using a shooting device. The image contains the original content displayed on the screen as well as various imaging features introduced during the shooting process.

[0018] Shooting equipment refers to electronic devices capable of capturing optical images and converting them into digital signals, including but not limited to smartphones, digital cameras, tablets, and other terminals with camera functions.

[0019] A screen refers to an output device used to display visual information, such as an LCD monitor, LED display, or OLED screen, whose surface has embedded hidden video watermark information.

[0020] Specifically, when a user needs to record sensitive information or protected content displayed on the screen, they will use a mobile phone or other camera device to take a picture or screenshot of the screen. In this scenario, the system first needs to receive and acquire this copied image generated through physical photography. This copied image can be acquired through various means, including but not limited to reading from the local storage of the camera device, receiving it via network transmission, or downloading it from a cloud storage service. Because the screen already embeds an invisible watermark when displaying content, this watermark information is captured by the camera of the camera device along with the light emitted from the screen, thus being preserved in the copied image.

[0021] In some embodiments, the acquisition of reproduced images can be achieved in multiple ways. Optionally, the system can establish an image upload interface through which users upload reproduced images stored in the shooting device to the server. After receiving the uploaded image data stream, the server first verifies the legality of the image format, checks whether the image is a common format such as JPEG or PNG, then caches the image data in a temporary storage area, and finally returns the image access path to the subsequent processing module, completing the process of acquiring reproduced images.

[0022] Optionally, the system can deploy a real-time monitoring program that runs on the shooting device. When it detects that the user has launched the camera application and is taking a picture of the screen, the monitoring program automatically intercepts the shooting completion event, immediately reads the newly generated image file, and transmits the image data to the watermark extraction system in real time through an encrypted channel. After receiving the image data, the system decrypts and verifies its integrity. Once it confirms that the image has not been tampered with, it stores it in a designated location, thereby completing the automated acquisition of the reproduced image.

[0023] S102: Obtain the screen parameters in the reproduced image, and divide the reproduced image into multiple grid units of equal size according to the screen parameters; Among them, screen parameters refer to a set of technical indicators that can characterize the physical characteristics and display performance of the screen, including but not limited to key information such as screen resolution, size, pixel density, color gamut, refresh rate, and panel type. These parameters determine the display effect and watermark embedding characteristics of the screen.

[0024] A grid cell represents a rectangular image patch of equal size and consistent shape obtained through regular segmentation. Each grid cell covers a local area of ​​the reproduced image and contains all pixel information within that area. Equal size means that all grid cells maintain complete consistency in pixel dimensions; for example, each grid cell is a 64×64 pixel or 128×128 pixel square area. This uniform segmentation is beneficial for subsequent batch processing and feature comparison.

[0025] Specifically, after successfully acquiring the reproduced image, the system needs to perform structured processing on the image in order to extract the watermark information.

[0026] First, the system uses image analysis technology to identify and extract screen parameters from the reproduced images. This process may include detecting screen boundaries in the image, analyzing the display content features of the screen, and identifying the physical size ratio of the screen. After obtaining the screen parameters, the system calculates the most suitable grid cell size based on these parameters. This size needs to match the screen resolution and the watermark embedding density to ensure that each grid cell can contain exactly one complete watermark cell or a fixed number of watermark pixels.

[0027] Subsequently, the system uses the effective display area of ​​the screen in the reproduced image as a benchmark, and uniformly divides the image according to the calculated grid cell size. Starting from the top left corner, grid cells are generated sequentially from left to right and from top to bottom until the entire screen area is covered. This gridded segmentation strategy transforms the complex global watermark extraction problem into multiple simple local watermark recognition tasks, which not only reduces computational complexity but also improves the accuracy of watermark localization. Especially when dealing with reproduced images containing geometric distortions or local blurring, the gridded method can effectively isolate damaged areas and prevent local problems from affecting the overall watermark extraction effect.

[0028] In some embodiments, image segmentation based on screen parameters can be implemented in various ways. Optionally, the system can first use an edge detection algorithm to identify the four boundary points of the screen in the photographed image, find the rectangular region of the screen through Hough transform or contour detection, then extract the width and height information of the rectangular region, calculate the standard resolution type of the screen based on the aspect ratio, then query the recommended grid size corresponding to the resolution from a pre-established screen parameter database, and finally segment the screen region according to the grid size to generate a grid matrix with a determined number of rows and columns, where each matrix element corresponds to a grid cell.

[0029] Optionally, the system can adopt an adaptive segmentation strategy. First, it analyzes the overall clarity and noise level of the photographed image. If the image quality is high, a smaller grid size is selected to obtain more precise watermark positioning. If the image quality is low, a larger grid size is selected to enhance robustness. Then, the spatial distribution density of the watermark is calculated based on the refresh rate and watermark frame rate information in the screen parameters. The size of the grid cells is dynamically adjusted in combination with the image quality assessment results. Finally, a grid partitioning scheme with a size adapted to the current image state is generated to complete the adaptive segmentation processing of the image.

[0030] S103: Perform feature analysis on the image of each grid cell and output the watermark recognition confidence score corresponding to each grid image cell; Watermark recognition confidence refers to the quantitative evaluation value of the system's certainty about the existence of watermark information in a certain grid cell. It is usually represented by a value between 0 and 1. The closer the value is to 1, the more confident the system is that the grid cell contains a valid watermark. The closer the value is to 0, the more likely it is that it does not contain a watermark or the watermark is seriously damaged.

[0031] Specifically, after completing the gridded segmentation of the reproduced image, the system needs to perform detailed feature analysis on each grid cell to determine whether it contains valid watermark information. This process first traverses all grid cells, extracting multi-dimensional image features for each cell. For example, it uses Fourier transform to convert the spatial domain image to the frequency domain and analyzes its spectral characteristics, since watermark information is usually embedded in specific frequency components. Simultaneously, the system also calculates the statistical characteristics of each grid cell, including pixel values ​​such as mean, variance, and entropy. These statistical characteristics reflect the complexity and regularity of the image.

[0032] In addition, the system extracts texture features, analyzing the texture structure of grid cells using gray-level co-occurrence matrix (GLCM) or local binary pattern (LOB) analysis, as watermark embedding alters the texture pattern of the original image to some extent. After extracting all features, the system comprehensively considers information from each feature dimension to calculate the probability that the grid cell contains the watermark; this probability value is the watermark recognition confidence score. For the confidence score calculation, the system may use the softmax probability value output by the classifier, or it may obtain it through a weighted average of the voting results of multiple weak classifiers, or it may perform normalization calculation based on the similarity between the feature vector and the standard watermark feature template.

[0033] S104: The grid cells with a watermark recognition confidence level greater than the preset confidence threshold are identified as target grid cells, and the feature patterns of the target grid cells are matched with the preset watermark template to determine the local watermark information of the target grid cells; The preset confidence threshold refers to a pre-set discrimination standard value used by the system to distinguish between high-confidence grid cells and low-confidence grid cells. Only grid cells with a confidence level exceeding the threshold are considered to reliably contain watermark information. This threshold is usually adjusted according to the accuracy requirements and fault tolerance requirements of the actual application scenario, for example, set to 0.7, 0.8 or 0.9.

[0034] The target grid cell represents the set of grid cells containing valid watermark information that have been determined after confidence screening. These cells are the focus of subsequent watermark extraction and tracing.

[0035] Feature patterns refer to specific patterns, encoding methods, or data distribution rules presented by watermark information in grid cells, such as a specific pixel arrangement, periodic changes at a specific frequency, or a specific encoding sequence. These patterns are the key basis for identifying and decoding watermarks.

[0036] The preset watermark template is used to represent the standard watermark feature reference library pre-stored by the system. It contains complete information such as feature patterns, encoding rules, and decoding algorithms of different types of watermarks, and is used to compare and match with the extracted features.

[0037] Local watermark information refers to partial watermark data extracted from a single or a few adjacent grid cells. Since the entire watermark information is distributed across multiple grid cells, each target grid cell can only provide a fragment or part of the complete watermark.

[0038] Specifically, after obtaining the confidence level of all grid cells, the system needs to filter out the reliable cells that actually contain the watermark and extract the watermark information from them.

[0039] First, the system compares the confidence level of each grid cell with a preset confidence threshold. This comparison filters out grid cells where the watermark information is damaged or unclear due to factors such as blurry images, uneven lighting, or screen reflections, retaining only cells with high confidence levels and clear watermark features as target grid cells. This filtering mechanism effectively improves the accuracy of watermark extraction and avoids interference from erroneous information.

[0040] After determining the target grid cell, the system performs more in-depth feature extraction on each target grid cell to obtain the specific feature pattern of the watermark in the cell. This pattern may be a string of binary code sequence, a parameter of a certain modulation signal, or a specific spatial distribution pattern.

[0041] Subsequently, the system matches the extracted feature patterns with various templates in the preset watermark template library. The matching process may employ various methods such as template matching algorithms, correlation analysis, and Hamming distance calculation. The goal is to find the watermark template most similar to the extracted features, thereby determining the encoding method used to embed the watermark in the target grid cell. Once a match is successful, the system can parse the feature patterns according to the corresponding decoding rules, extracting the local watermark information contained in the target grid cell. This information may include a few characters of the user ID, certain bits of the timestamp, or partial data of the device identifier.

[0042] It is important to note that since watermarking typically uses a distributed embedding method, a single grid cell often cannot provide complete watermark content. Therefore, this step extracts local watermark information, which needs to be fused with the local information of other target grid cells in subsequent steps to restore the complete watermark data.

[0043] S105: Based on the splicing position obtained after merging and splicing multiple adjacent target grid cells, multiple local watermark information is fused to obtain global watermark information; Merging and stitching refers to the operation of recombining multiple scattered target grid units into a continuous whole region according to their relative positional relationship in the original image. This process requires maintaining the spatial order and positional relationship between the units.

[0044] The splicing position is used to indicate the specific coordinate position of each target grid cell in the overall area formed after merging, including the index number of the cell in the row and column directions. This positional information is the key to correctly restoring the order of the watermark data.

[0045] Global watermark information refers to the complete watermark data obtained after fusion processing. It contains all traceability information embedded in the entire screen display content, such as complete user identifiers, timestamps, device numbers, IP addresses, and other key data that can be used for traceability.

[0046] Specifically, after extracting local watermark information from each target grid cell, the system needs to integrate these scattered information fragments into complete watermark content. This process first requires determining the spatial relationships between the target grid cells. By recording the row and column indices of each target grid cell in the original grid matrix, the system can accurately determine which cells are adjacent and which cells have spatial intervals.

[0047] Next, the system identifies all adjacent target grid cell groups and virtually stitches these adjacent cells together according to their original positional relationships to form one or more continuous target regions. Each target region may correspond to a complete paragraph or section of watermark information. During the stitching process, the system assigns a stitching position number to each target grid cell. This number reflects the relative order of the cell in the stitched region, such as the cell in the first row and second column, or the cell in the third row and first column. With this stitching position information, the system can determine the correct arrangement order and logical relationship between the various local watermark information.

[0048] Subsequently, the system arranges the local watermark information in an orderly manner according to the splicing position, connecting these pieces of information sequentially from left to right, from top to bottom, or in other predefined orders to form preliminary global watermark information. During the connection process, the system also needs to handle potential information redundancy, overlap, or missing information. For example, some adjacent units may contain partially duplicated watermark data for error correction. The system needs to identify these redundancies and remove duplicates. For missing parts, the system can recover them through error correction coding or infer and complete them based on contextual information, ultimately obtaining complete and accurate global watermark information. This information contains all the necessary data for traceability and can clearly identify the source of the reproduced image and related shooting information.

[0049] S106: Obtain the camera imaging parameters of the shooting device, and optimize the global watermark information according to the camera imaging parameters to obtain the target watermark information of the reproduced image. The target watermark information is used to trace the reproduction information of the reproduced image.

[0050] Among them, camera imaging parameters refer to various technical parameters and performance indicators involved in the image acquisition and processing of the camera of the shooting device, including sensor type, pixel size, lens focal length, aperture size, ISO sensitivity, white balance setting, image processing algorithm type, etc. These parameters together determine the imaging quality and features of the final generated image.

[0051] The target watermark information refers to the final watermark result obtained after optimization processing. This result has eliminated the influence of shooting equipment and shooting environment as much as possible, and can accurately and reliably reflect the original embedded watermark content.

[0052] Source tracing refers to the process of tracing and locating the source of a photographed image using extracted watermark information. By analyzing the identification information contained in the watermark, it is possible to determine which user, at what time, and using what device captured the screen content.

[0053] Reproduction information refers to various metadata related to the reproduction behavior, including key information that can be used for tracing and evidence collection, such as the identity of the photographer, the shooting time, the shooting location, the shooting equipment model, and the original screen source.

[0054] Specifically, after obtaining the global watermark information, the system also needs to consider the impact of the shooting device's characteristics on the watermark extraction results and perform targeted optimizations. Different brands and models of shooting devices use different image sensors, lens components, and image processing chips, resulting in different imaging effects when shooting the same screen. For example, some mobile phones may aggressively sharpen the image, leading to enhanced or distorted watermark features; some cameras may apply strong noise reduction algorithms, causing weak watermark signals to be filtered out; and some devices may have severe color deviations, causing decoding errors in color-coded watermark information. Therefore, the system first needs to obtain the camera imaging parameters of the shooting device. These parameters can be obtained through various means, such as reading the device model and shooting parameters from the image's EXIF ​​metadata, or inferring the device type by analyzing the statistical features of the image using device fingerprint recognition technology.

[0055] After acquiring the imaging parameters, the system queries the imaging characteristic model of the device from a pre-established device characteristic database based on these parameters. This model describes the various systematic deviations and noise characteristics introduced by the device during image acquisition and processing.

[0056] Next, the system performs reverse compensation processing on the global watermark information based on the imaging characteristic model. For example, if the device tends to over-sharpen, a smoothing filter is applied to the watermark information for anti-sharpening; if the device has a fixed color shift, color correction is applied to the watermark data in the opposite direction; if the device's sensor noise is high, the signal-to-noise ratio of the watermark signal is enhanced or a stronger error correction algorithm is applied. Through these targeted optimization processes, the system can effectively eliminate the negative impact of the shooting device characteristics on watermark extraction, obtaining more accurate and reliable target watermark information. This target watermark information contains complete and accurate traceability data, which can be used for subsequent traceability analysis. For example, by parsing the user ID field in the watermark, it is possible to find out which account viewed the content; the timestamp field can determine the specific time the shooting occurred; and the device identification field can pinpoint the specific terminal used for shooting. This information together constitutes a complete chain of evidence for the re-filming behavior, providing a reliable basis for the investigation of screen information leakage.

[0057] Based on the above embodiments, as an optional implementation method, the method of dividing the photographed image into multiple grid units of equal size according to the screen parameters in S102 can be implemented through the following steps S201-S204.

[0058] S201: Identify the screen boundaries in the photographed image and determine the effective display area of ​​the screen; Among them, the screen boundary refers to the dividing line between the screen display panel and the surrounding non-display area in the reproduced image. This boundary is usually formed by features such as the screen border, the brightness difference between the screen and the background, and the color contrast between the displayed content and the non-display area.

[0059] The effective display area refers to the rectangular area within the screen boundary that is actually used to display content. This area includes the complete display image and the watermark information embedded within it, excluding irrelevant parts such as screen borders, shooting background, and reflective areas.

[0060] After acquiring the reproduced image, the image is first preprocessed to reduce noise and improve contrast. Then, an edge detection algorithm is used to extract the edges of the entire image. This algorithm identifies the locations in the image where there are significant changes in brightness or color by calculating pixel gradient changes. There are usually obvious brightness jumps at the screen boundaries, which will produce a strong edge response.

[0061] Next, straight line fitting is performed on the detected edges. Since screen boundaries typically consist of four mutually perpendicular straight line segments, the straight line detection algorithm can filter out candidate boundary lines that conform to this geometric feature from numerous edges. Subsequently, the candidate boundary lines are verified and optimized by analyzing features such as brightness distribution, color statistics, and texture regularity within the boundary line area to determine whether the region is the actual screen display area. After confirming the screen boundary, the rectangular area enclosed by the four boundary lines is calculated; this rectangle is the effective display area, and parameters such as the top-left corner coordinates, width, and height of this area are recorded.

[0062] S202: Match the corresponding screen model from the preset screen parameter database according to the screen parameters, and determine the display characteristic parameters corresponding to the screen model; The screen parameter database refers to a pre-established and stored structured dataset containing various screen models and their corresponding technical specifications. This database records detailed parameter information such as resolution, pixel density, refresh rate, panel type, and color gamut range for different brands and models of screens.

[0063] The screen model refers to the identifier of a specific screen product, which corresponds to a set of specific hardware specifications and display performance indicators. Display characteristic parameters refer to a set of technical indicators describing the screen's display performance and imaging characteristics, including key characteristics that affect watermark embedding and extraction, such as pixel arrangement, sub-pixel structure, brightness uniformity, response time, and color calibration curve.

[0064] Specifically, starting with the size information of the effective display area, the aspect ratio of this area is calculated, reflecting the screen's aspect ratio characteristics. Simultaneously, the pixel distribution density within the effective display area is analyzed. By detecting spatial correlation and periodic patterns between adjacent pixels, the pixel arrangement structure of the screen is inferred. These extracted parameters are used as query conditions and retrieved from a pre-defined screen parameter database. Each record in the database contains standard parameter values ​​for a specific screen model. By calculating the degree of difference between the query parameters and the parameters recorded in the database, a matching score is generated for each candidate screen model.

[0065] The matching score comprehensively considers multiple dimensions such as the degree of aspect ratio matching, resolution similarity, and pixel density consistency. The higher the score, the more similar the screen model is to the actual screen in the photograph. The database record with the highest matching score is selected, and its corresponding screen model is determined as the screen model in the currently reproduced image. At the same time, the complete display characteristic parameters of this model stored in the database are obtained.

[0066] S203: Determine the grid unit size that matches the screen resolution based on the display characteristic parameters, and uniformly divide the effective display area according to the grid unit size to obtain multiple grid units of equal size.

[0067] Resolution refers to the number of pixels a screen can display in the horizontal and vertical directions. It is usually expressed as the number of horizontal pixels multiplied by the number of vertical pixels. Resolution directly determines the level of detail and information capacity of the screen display.

[0068] The grid cell size refers to the number of pixels contained in a single grid cell in both the width and height directions. This size parameter determines how many rows and columns the effective display area is divided into.

[0069] Specifically, after obtaining the screen's display characteristic parameters, the appropriate grid cell size is calculated based on the resolution information. The calculation process needs to comprehensively consider factors such as the total number of pixels on the screen, the watermark embedding density, and the minimum data volume required for subsequent feature analysis. Based on the watermark embedding scheme recorded in the display characteristic parameters, the pixel range occupied by each watermark cell on the screen is determined. The grid cell size needs to be set to an integer multiple of the watermark cell size to ensure that each grid cell contains exactly the full number of watermark cells, avoiding watermark information being truncated by grid boundaries and causing extraction failure.

[0070] At the same time, the scaling relationship between the actual resolution of the reproduced image and the original screen resolution should be considered. Because the number of pixels in the effective display area of ​​the reproduced image is usually different from the actual screen resolution due to the shooting distance and lens focal length during the shooting process, the theoretical grid unit size needs to be adjusted according to the scaling ratio to obtain the actual grid unit size applicable to the current reproduced image.

[0071] After determining the grid cell size, starting from the top left corner of the effective display area, the grid boundaries are sequentially divided horizontally according to this size. When the right boundary is reached, the division continues on the next line until the entire effective display area is covered. During the division process, the size consistency of each grid cell is strictly maintained. For edge locations where the number of pixels is insufficient to form a complete grid cell, appropriate measures are taken: either these incomplete edge parts are discarded, or the size of the grid cells in the last row and last column is fine-tuned to completely cover the effective area. The resulting multiple grid cells form a two-dimensional matrix. The number of rows in the matrix equals the height of the effective display area divided by the grid cell height, and the number of columns equals the width of the effective display area divided by the grid cell width. Each grid cell corresponds to an element position in the matrix and contains the image data at that position.

[0072] Based on the above embodiments, as an optional implementation method, the construction of the screen parameter database can be specifically achieved through the following steps S301-S304.

[0073] S301: Collect parameter information of various brands, models or specifications of screens. The parameter information includes at least one of the following: screen resolution, color space type, brightness range, contrast range, refresh rate and screen display technology type. Among them, parameter information refers to technical indicator data that describes the screen hardware specifications and display performance. These data record the screen's physical characteristics and display capabilities in the form of numerical values, categories, or ranges.

[0074] Color space type refers to the range of colors a screen can display and the color coding standard. Different types use different color representation methods and color gamut coverage. Common types include standard red-green-blue color space and wide color gamut color space.

[0075] Brightness range refers to the interval between the minimum and maximum brightness values ​​that a screen can output. This range reflects the screen's ability to display both light and dark areas and its dynamic range.

[0076] Contrast range refers to the range of brightness ratios between the brightest white and the darkest black displayed on a screen. This parameter reflects the screen's ability to distinguish between different levels of brightness and darkness.

[0077] Refresh rate refers to the number of times the screen updates its display per second. This frequency affects the smoothness of dynamic images and the display response speed.

[0078] Screen display technology type refers to the imaging technology principle and panel structure category used in the screen. Different technology types have fundamental differences in terms of light emission mode, pixel control method, color performance, etc.

[0079] Specifically, to establish a parameter database covering various screen types, technical specifications for different screens were first obtained through multiple channels. Officially labeled technical parameters were extracted from product specification sheets released by screen manufacturers; these parameters, verified by the manufacturers, possess high accuracy and authority. Actual performance data measured with professional instruments was obtained from screen evaluation reports published by industry testing organizations; this data, derived through standardized testing procedures, reflects the screen's true performance. All collected screen parameter information was then standardized and organized, unifying the unit formats and representations of data from different sources. Parameter data for the same screen model obtained from multiple channels was cross-validated, eliminating obviously abnormal values ​​and retaining the most reliable ones, thus forming a standardized set of screen parameter information.

[0080] S302: Determine the display characteristic simulation parameters of each screen based on the parameter information of each screen, and calculate the display influence coefficient of each screen on the embedded watermark based on the display characteristic simulation parameters. The display influence coefficient represents the degree of difference between the image before and after the watermark is embedded. Among them, display characteristic simulation parameters refer to a set of technical parameters used to simulate the actual display effect of the screen in a computer. These parameters are based on the physical characteristics of the screen and are obtained through mathematical model conversion. They can reproduce the display behavior of the screen, such as color mapping, brightness modulation, and contrast adjustment, at the software level.

[0081] The display impact coefficient is a numerical indicator that quantifies the degree of change that the screen display process brings to the embedded watermark. This coefficient is calculated by comparing the difference between the original image data before the watermark is embedded and the image data after it has been displayed on the screen. The larger the value, the more significant the screen's impact on the watermark.

[0082] Image difference refers to the degree of difference between two images in terms of pixel values, color distribution, brightness levels, frequency domain characteristics, etc. This difference is measured by quantitative analysis methods and converted into comparable values.

[0083] Specifically, firstly, a color conversion matrix is ​​determined based on the color space type of each screen. This matrix describes the mapping relationship from standard color encoding to the actual output color of the screen. A brightness mapping curve is established based on the brightness range parameter. This curve defines the correspondence between the input brightness value and the actual displayed brightness value on the screen. The shape of the mapping curve varies between different screens; some screens use linear mapping, while others use non-linear gamma correction curves. A contrast adjustment function is set based on the contrast range parameter. This function adjusts the contrast intensity of bright and dark parts in the image. Screens with a larger contrast range can retain more details in both bright and dark areas, while screens with a smaller contrast range will compress the tonal gradations. A temporal response characteristic parameter is determined based on the refresh rate and display technology type. This parameter describes the time delay and transition process of changes in screen pixel values, affecting the display quality of dynamic content. Combining the color conversion matrix, brightness mapping curve, contrast adjustment function, and temporal response characteristic parameter constitutes the set of simulated display characteristic parameters for the screen.

[0084] Next, the display impact coefficient is calculated. A set of standard test images containing watermarks of varying intensities is selected. The watermark embedding position, embedding strength, and encoding method of these images are known and fixed. The original data of each test image is used as input, and display characteristic simulation parameters are applied to process the image, simulating the image data obtained by photographing the image after it has been displayed on the current screen. The image data before and after processing are compared, the average deviation of pixel values ​​is calculated, the offset direction and magnitude of color components are analyzed, the energy change of the image spectrum is measured, and the differences in changes between the watermarked and non-watermarked areas are statistically analyzed. These difference indicators are then comprehensively weighted to obtain a value characterizing the degree of influence of the screen on the watermark; this value is the display impact coefficient.

[0085] The display influence coefficient is calculated for all images in the test image set, and the average value is taken as the final display influence coefficient for that screen. The above process is repeated for each collected screen to determine its display characteristic simulation parameters and display influence coefficient. These data will be used to guide the adaptive adjustment of watermark embedding strength and the parameter configuration of the watermark extraction algorithm.

[0086] S303: Based on the numerical range of resolution of each screen, the corresponding screens are divided into multiple resolution levels, and each resolution level is set with a corresponding grid unit size parameter. Among them, resolution level refers to the classification level based on the size range of screen resolution values. Screens in the same level have resolution values ​​that fall within the same range and have similar pixel density and display sharpness.

[0087] Specifically, after determining the display characteristics of each screen, appropriate grid division standards need to be set for screens with different resolutions. First, all collected screens are sorted according to their resolution values. The resolution value is obtained by multiplying the horizontal and vertical pixel counts to obtain the total pixel count, which comprehensively reflects the screen's display capacity. The distribution of resolutions after sorting is observed, identifying dense and sparse intervals in the value distribution, and setting level division boundaries at natural breakpoints in the distribution. Based on the standard specifications of common screen resolutions and the distribution characteristics of mainstream products in the market, the entire resolution range is divided into several non-overlapping level intervals, each interval corresponding to a resolution level. Each resolution level has a lower limit and an upper limit. When a screen's resolution value falls between the lower and upper limits of a level, the screen belongs to that resolution level. For each divided resolution level, the corresponding grid unit size parameters are determined based on the typical resolution value of that level. For example, standard definition, high definition, and ultra-high definition are graded.

[0088] S304: Establish and store the correspondence between the parameter information, display influence coefficient and grid cell size parameters of each screen, and build a screen parameter database.

[0089] Specifically, the first step is to design the logical structure of the database, determining the field settings of the data tables and the relationships between tables. A basic screen information table is created, using the screen model as the primary key. This table records identification information for each screen, such as brand, model, and release year, as well as parameter information fields such as resolution, color space type, brightness range, contrast range, refresh rate, and display technology type. A display characteristics table is created, also using the screen model as the association key. This table stores the simulated display characteristics parameters and display influence coefficients for each screen. The simulated display characteristics parameters are stored as parameter groups or parameter file paths, and the display influence coefficients are stored as numerical fields. A resolution level table is created, recording the number, lower limit, upper limit, and grid cell size parameters for each resolution level, with each level occupying one record.

[0090] A resolution level field is added to the screen basic information table. Based on the resolution value of each screen, the corresponding level number is looked up in the resolution level table and entered into the resolution level field of the screen basic information table, thus establishing a relationship between the screen and the resolution level. Through this relationship, when information about a screen is retrieved, both the resolution level to which the screen belongs and the corresponding grid cell size parameters can be obtained simultaneously. The compiled screen data is then inserted into the corresponding tables in the database one by one, ensuring that the parameter information, display impact coefficients, and resolution level associations for each screen are accurate. Indexes are created for key fields in the database tables, including screen model indexes, resolution value indexes, and resolution level indexes. These indexes accelerate the data query process; when screen parameters are input for matching, the system can quickly locate the most similar screen record and return its complete information.

[0091] Based on the above embodiments, as an optional implementation method, the feature analysis of the image of each grid unit in S103 and the output of the watermark recognition confidence level corresponding to each grid image unit can be implemented through the following steps S401-S404.

[0092] S401: Perform feature analysis on the image of each grid cell and extract the image features of each grid cell; Image features refer to quantitative data that characterize the visual attributes and content characteristics of an image. These data include various forms such as color statistics describing color distribution, texture patterns describing spatial structure, geometric shapes describing edge contours, and frequency domain coefficients describing frequency components. Through these features, the core information of an image can be grasped without directly processing all pixels.

[0093] Specifically, firstly, for a single grid cell, the color values ​​of all pixels within that cell are read. These color values ​​are typically stored in the form of red, green, and blue three-channel components. Statistical analysis is then performed on these pixel values ​​to calculate the average value of each color channel within the grid cell. This average value reflects the overall tonal tendency of the cell. The variance of each color channel pixel value is then calculated, reflecting the dispersion and fluctuation of the color distribution within the cell. Finally, a frequency domain transformation is performed on the grid cell, converting the spatial domain pixel distribution into a frequency domain coefficient distribution. Frequency domain coefficients reveal the intensity of different spatial frequency components in the image. Since watermark information is usually embedded within a specific frequency range, frequency domain coefficients are a crucial basis for watermark detection.

[0094] Texture features of grid cells are extracted by analyzing spatial correlations and gray-level co-occurrence patterns between pixels to obtain feature parameters describing the texture coarseness, directionality, and regularity within the cell. Edge features are extracted by using edge detection operators to identify locations within the grid cell where brightness or color changes drastically, and the number, direction distribution, and intensity distribution of edges are statistically analyzed. These edge features reflect the structural complexity of the cell content. All of these feature parameters are organized into a feature vector, which comprehensively describes the image features of the grid cell.

[0095] S402: Identify the watermark quality features in the image features of each grid cell and determine the initial confidence value of each grid cell; Among them, watermark quality features refer to specific feature components in image features that are directly related to the watermark embedding and display quality. These features can indicate the watermark's presence intensity, clarity, distortion level, and other quality attributes within the grid cell.

[0096] The initial confidence value is a quantitative value that independently evaluates the reliability of a grid cell containing effective watermark information based on the watermark quality characteristics of the individual grid cell itself. The higher the value, the more obvious the watermark features in the cell, the better the quality, and the greater the probability of successful recognition.

[0097] Specifically, the watermark information is first represented in the image features according to the specific scheme used for watermark embedding. When the watermark is embedded in the frequency domain coefficients, coefficient components corresponding to a specific frequency range in the feature vector are selected as watermark quality features. The amplitude distribution and phase relationship of these coefficients are analyzed to detect whether there is a coefficient modulation pattern that conforms to the watermark coding rules.

[0098] When a watermark is embedded in a color channel, color statistical features are selected as watermark quality features to check whether there are expected correlation changes between color channels or whether there are small numerical shifts within a specific channel. Signal-to-noise ratio (SNR) analysis is performed on the extracted watermark quality features, calculating the ratio of the intensity of the watermark signal component to the intensity of the background noise component. A high SNR indicates that the watermark features are prominent and easy to identify, while a low SNR indicates that the watermark features are masked by noise and difficult to identify.

[0099] Based on the theoretical model of watermark encoding, an ideal watermark feature template that the grid cell should possess is generated. The actual extracted watermark quality features are compared with the ideal template, and the matching degree between the two is calculated. A high matching degree indicates that the watermark of the cell is well preserved, while a low matching degree indicates that the watermark has been damaged or interfered with. Taking into account both the signal-to-noise ratio and the template matching degree, an initial confidence value is assigned to the grid cell. This value is obtained by combining the signal-to-noise ratio and the matching degree according to a certain weight. The above watermark quality feature identification and initial confidence calculation process is performed on all grid cells, and each grid cell obtains an independent initial confidence value.

[0100] Based on the above embodiments, as an optional implementation method, the method of identifying the watermark quality features in the image features of each grid cell and determining the initial confidence value of each grid cell in S402 can be specifically implemented through the following steps S4021-S4022.

[0101] S4021: Compare the image features of each grid cell with the preset watermark quality feature template, identify the feature parts of the image features of each grid cell that conform to the preset watermark quality feature template, and use them as the watermark quality features of each grid cell. Among them, the preset watermark quality feature template refers to the standard feature pattern established in advance for identifying and verifying watermark information. The template contains standardized descriptive information such as the typical feature parameter range, feature distribution law, frequency domain coefficient pattern or color modulation pattern that the watermark should present in the image, which serves as a reference benchmark for judging which parts of the image features belong to the watermark features.

[0102] Specifically, after acquiring the complete image features of each grid cell, it is necessary to accurately identify the feature components that specifically reflect the watermark quality from these comprehensive features. First, the details of the preset watermark quality feature template are read. This template is pre-constructed according to the design rules of the watermark embedding algorithm and records the ideal representation of the watermark information in the image feature space. For the currently processed grid cell, the complete image feature vector obtained in the feature extraction stage is acquired. This vector contains feature components of multiple dimensions, including color statistical features, texture features, frequency domain features, and edge features. Based on the watermark embedding location information specified in the preset watermark quality feature template, the corresponding feature dimension in the image feature vector is located.

[0103] When a watermark is embedded in the frequency domain, the frequency domain transform coefficients are extracted from the image feature vector. This part records the energy distribution of the grid cells at different frequency components. The extracted frequency domain coefficients are compared with the watermark frequency band range specified in the template. The template specifies the frequency range in which watermark information is typically embedded, and frequency domain coefficients falling within this range are selected. The selected frequency domain coefficients are further examined for their numerical distribution patterns. The template specifies the coefficient modulation rules caused by watermark encoding, and the actual coefficients are checked to see if they exhibit this modulation rule.

[0104] When a watermark is embedded in a color channel, statistical parameters for each color channel are extracted from the image feature vector. The mean and variance of each channel are compared with the color shift pattern caused by the watermark as specified in the template to identify channel parameters exhibiting specific shift patterns. For texture features, when the template indicates that the watermark will change certain texture properties, relevant parameters in the texture features are extracted, and it is checked whether these parameters deviate from the typical texture pattern of a natural image, thus revealing traces of watermark embedding. All feature components confirmed through template comparison are summarized, and these components constitute the watermark quality features of that grid cell.

[0105] The watermark quality features are a subset of the original image features, specifically containing watermark-related features while removing background image features unrelated to the watermark. The template comparison and feature recognition process described above is repeated for each grid cell, resulting in individual watermark quality features for all grid cells.

[0106] S4022 calculates the initial confidence value of each grid cell based on the degree of matching between the watermark quality characteristics of each grid cell and the preset watermark feature template; Specifically, the ideal watermark feature parameters described in the preset watermark feature template are first read. This template is based on the theoretical design of the watermark coding algorithm and reflects the standard form that watermark features should present under conditions of no damage and no interference. For the watermark quality features extracted from the current grid cell, a dimension-by-dimensional comparative analysis is performed with the ideal watermark feature template. For frequency domain watermark quality features, the actual frequency domain coefficients of the grid cell within the watermark frequency band are extracted, the theoretical coefficient values ​​for the corresponding frequency band in the template are read, and the degree of difference between the actual and theoretical coefficients is calculated.

[0107] The degree of difference is quantified by calculating the Euclidean distance between two coefficient vectors. The smaller the distance, the closer the actual watermark frequency domain features are to the theoretical standard, and the higher the degree of matching. For color modulation type watermark quality features, the actual modulation amount of each color channel of the grid cell is extracted, the ideal modulation amount specified in the template is read, and the deviation between the actual modulation amount and the ideal modulation amount is calculated. The deviation is quantified by calculating the relative error between the two. The smaller the relative error, the more accurate the color modulation and the higher the degree of matching. For encoded sequence type watermark quality features, the actual bit sequence decoded from the grid cell is extracted, the standard bit sequence stored in the template is read, and the consistency of the two sequences is compared bit by bit. The consistency is quantified by calculating the proportion of bits with the same value at the same position in the two sequences to the total number of bits. This proportion is the degree of matching. After obtaining the degree of matching in each feature dimension, the degree of matching in these sub-dimensions is integrated.

[0108] When the watermark quality features include multiple dimensions, a weighted average of the matching degree of all dimensions is calculated. The weights are assigned according to the importance of each dimension to watermark recognition, with more important dimensions receiving higher weights. The weighted average yields the overall matching degree of the watermark quality features for that grid cell, which directly reflects the integrity of the watermark preservation and the reliability of recognition within that cell.

[0109] The overall matching degree is used as the initial confidence value for this grid cell, or the overall matching degree is converted into an initial confidence value through a mapping function. The mapping function maps the matching degree from the original numerical range to a standardized confidence range, typically set between zero and one. The confidence value is one when the matching degree is highest and zero when the matching degree is lowest. Intermediate matching degrees are mapped to intermediate confidence values ​​according to a linear or non-linear relationship. The above matching degree calculation and initial confidence assignment process is performed on all grid cells, and each grid cell obtains an initial confidence value based on the matching degree between its watermark quality features and the ideal template.

[0110] S403: For each grid cell, obtain the watermark candidate features of the grid cells that are spatially adjacent to the grid cell, and calculate the similarity between the watermark candidate features of the grid cell and the watermark candidate features of the adjacent grid cells; Among them, watermark candidate features refer to the potential watermark information content decoded from the image features of grid cells. These features exist in the form of binary sequences, numerical codes, or pattern identifiers, representing the watermark data embedded in the grid cell.

[0111] Similarity is a quantitative measure of the degree of consistency between two watermark candidate features. It is calculated by comparing the numerical differences, pattern matching degree, or encoding distance between the two features. High similarity indicates that the contents of the two features are close or the same, while low similarity indicates that the contents of the two features are significantly different.

[0112] Specifically, watermarks typically employ the same or related encoding information in adjacent regions during embedding; therefore, the watermark features of adjacent grid cells should exhibit a certain degree of similarity or continuity. For the currently processed grid cell, its row and column position in the grid matrix is ​​first determined, and all directly adjacent grid cells are identified based on this position. For cells located at grid edges, the number of their adjacent cells is less than eight, including only the actually existing adjacent cells.

[0113] The process involves acquiring the watermark candidate features for the current grid cell. These features are obtained through decoding or pattern recognition of the watermark quality features for that cell. The decoding process converts frequency domain coefficients or color modulation information into the original watermark bit sequence or identifier based on the watermark encoding scheme. Similarly, the watermark candidate features for each adjacent grid cell are acquired, using the same decoding method to ensure feature comparability. The similarity between the watermark candidate features of the current grid cell and those of each adjacent cell is calculated. When the watermark candidate features are binary sequences, the similarity is determined by counting the number of bits with the same value at corresponding positions in the two sequences, with the proportion of identical bits to the total number of bits. When the watermark candidate features are numerical codes, the difference between the two values ​​is calculated; the smaller the difference, the higher the similarity. A standardized similarity value is obtained by mapping the difference to the range of zero to one. When the watermark candidate features are pattern identifiers, it is determined whether the two identifiers are completely identical. If they are identical, the similarity is one; otherwise, the similarity is zero.

[0114] S404: Adjust the initial confidence level based on the similarity to obtain the watermark recognition confidence level corresponding to each network unit.

[0115] Specifically, after obtaining the similarity information between each grid cell and its neighboring cells, the initial confidence level is optimized and adjusted using this spatial correlation information. The basic principle of the adjustment is that when the watermark candidate features of a grid cell are highly similar to the watermark candidate features of most of its neighboring cells, it indicates that the watermark recognition result of that cell has achieved spatial mutual verification, and its confidence level should be increased. Conversely, when the watermark candidate features of a grid cell are generally different from those of its neighboring cells, it indicates that the recognition result of that cell is isolated, and its confidence level should be decreased.

[0116] For the current grid cell, its initial confidence value and similarity values ​​with each neighboring cell are read. The average of all similarity values ​​is calculated, reflecting the overall consistency level between the cell and its surrounding environment. When the average similarity is high, a positive adjustment is added to the initial confidence value. The magnitude of the adjustment is proportional to the average similarity; the higher the average similarity, the larger the adjustment. When the average similarity is low, a negative adjustment is applied to the initial confidence value. The magnitude of the adjustment is inversely proportional to the average similarity; the lower the average similarity, the larger the adjustment.

[0117] In addition to considering the average similarity, the distribution pattern of similarity is also analyzed. When the similarity values ​​of all adjacent cells are relatively close, it indicates good consistency in the surrounding environment, thus increasing the confidence level. When the similarity values ​​of adjacent cells are discrete and differ significantly, it indicates uncertainty in the surrounding environment, thus appropriately reducing the confidence level. For grid cells with high initial confidence and high average similarity, the adjusted confidence level is close to the maximum value; these cells are considered to contain high-quality and spatially consistent watermark information. For grid cells with low initial confidence and low average similarity, the adjusted confidence level is further reduced; these cells may not contain a valid watermark or the watermark may be severely damaged.

[0118] Based on the above embodiments, as an optional implementation method, in S105, multiple local watermark information is fused according to the splicing position obtained after merging and splicing multiple adjacent target grid cells to obtain global watermark information. This can be specifically achieved through the following steps S501-S505.

[0119] S501: Obtain the spatial coordinate information of multiple adjacent target grid cells in the photographed image, and determine the relative positional relationship between each target grid cell; Spatial coordinate information refers to the positional description data of the grid cell in the two-dimensional plane coordinate system of the reproduced image. It usually includes geometric parameters such as the x and y coordinates of the upper left corner vertex of the grid cell, the width and height of the grid cell, etc. These parameters completely define the rectangular area occupied by the grid cell in the image.

[0120] Relative positional relationship refers to the geometric association between multiple target mesh elements in spatial distribution, including spatial topological information such as adjacent directions, spacing, and arrangement order between elements. This relationship describes the relative positional configuration of each element in the overall layout.

[0121] Specifically, the spatial coordinate information of each target grid cell is extracted from the location data stored during the grid division stage. For regular rectangular grid division, each grid cell is assigned a row index and a column index during division. Based on the row and column indices, as well as the number of rows and columns of the grid, the x-coordinate and y-coordinate positions of the top-left vertex of the grid cell in the coordinate system of the reproduced image are calculated. The x-coordinate is obtained by multiplying the column index by the width of a single grid cell, and the y-coordinate is obtained by multiplying the row index by the height of a single grid cell. The width and height dimensions of the grid cell are recorded; these dimensions are consistent across all cells in the regular grid. Combining the top-left corner coordinates, width, and height completely describes the rectangular area occupied by the target grid cell in the reproduced image. After obtaining the spatial coordinate information of all target grid cells, the relative positional relationships between them are analyzed. The coordinate information of any two target grid cells is compared one by one, and their row and column indices are compared to determine whether they are adjacent.

[0122] When two cells have the same row index and differ by one in their column index, they are determined to be left-right adjacent, with the cell with the smaller column index on the left and the cell with the larger column index on the right. When two cells have the same column index and differ by one in their row index, they are determined to be top-bottom adjacent, with the cell with the smaller row index on top and the cell with the larger row index on the bottom. When both the row index and column index of two cells differ by one, they are determined to be diagonally adjacent. For non-adjacent target grid cells, the distance between them is calculated, represented by a combination of the row index difference and the column index difference.

[0123] S502: Merge and stitch multiple adjacent target mesh cells according to their relative positional relationship to obtain the stitching position of the target mesh cells in the stitched target area; Specifically, a local coordinate system is established with the top-left corner of the target area as the origin. This coordinate system is independent of the global coordinate system of the reproduced image and is specifically used to describe the positional relationships within the target area. For each target grid cell, its coordinates in the global coordinate system of the reproduced image are converted to coordinates in the local coordinate system of the target area. The conversion method is to subtract the x-coordinate of the left boundary of the target area from the x-coordinate of the top-left corner of the cell in the global coordinate system to obtain the x-coordinate of the cell in the target area coordinate system, and subtract the y-coordinate of the top boundary of the target area from the y-coordinate of the top-left corner of the cell in the global coordinate system to obtain the y-coordinate of the cell in the target area coordinate system. After the coordinate transformation, each target grid cell obtains its stitching position within the target area. This stitching position describes the specific placement of the cell in the integrated target area, maintaining the original relative positional relationship between the cell and other cells.

[0124] S503: Obtain the information content of the local watermark information corresponding to each target grid cell; The information content refers to the actual data content carried by the local watermark information, including specific bit values, character encoding, pattern pixel values, or other forms of information payload. This content is a component of the effective information that the watermark is meant to convey.

[0125] Specifically, after determining the splicing positions of each target grid cell, it is necessary to extract the local watermark information content carried by each cell. For each target grid cell, the watermark quality feature data obtained during the watermark quality feature recognition stage is read. According to the specific design of the watermark coding scheme, the watermark quality features are decoded to convert the values ​​or patterns in the feature space into the original information content. When the watermark uses frequency domain spread spectrum coding, the frequency domain coefficients of the grid cell within the watermark frequency band are extracted, and these coefficients are correlated with a preset spreading code sequence. By detecting the position and amplitude of the correlation peaks, the information bits carried by the cell are demodulated.

[0126] When the watermark uses the least significant bit embedding method, the color value of the pixel within the grid cell is extracted, and the lowest few bits of the color channel value of each pixel are read. These lowest bits are then combined into a bit sequence according to the pixel scanning order. Error correction processing is performed on the decoded local watermark information, using the checksum and redundancy information of the error correction code to detect and correct erroneous bits in the local information. The decoding and error correction processes are performed on all target grid cells, resulting in a local watermark information content for each target grid cell.

[0127] S504: Determine the correlation between local watermark information based on the splicing position of each target grid cell; Specifically, after acquiring the local watermark information of each target grid cell, it is necessary to clarify how these scattered local information fragments are related to each other to form a complete watermark. The relationship between local watermark information is mainly determined by the splicing position of each target grid cell in the target area, and the spatial order of the splicing position corresponds to the logical combination order of the local watermark information. First, based on the global structure design of watermark encoding, the spatial distribution rules of the complete watermark information are determined. When the complete watermark information is encoded in the form of a linear sequence, each local watermark information is arranged and combined sequentially from left to right and from top to bottom in a spatial scanning order. For all target grid cells, they are sorted according to their splicing position coordinates in the target area. The sorting rule is to first sort by the vertical coordinate from smallest to largest, and for cells with the same vertical coordinate, then sort by the horizontal coordinate from smallest to largest, resulting in a target grid cell sequence that reflects the spatial scanning order.

[0128] When the complete watermark information is encoded in a two-dimensional matrix, each local watermark information maintains its two-dimensional spatial distribution relationship within the target area. For each target grid cell, its splicing position coordinates directly correspond to the row and column positions of the local watermark information in the global watermark matrix. A complete association mapping table is established for all target grid cells and their corresponding local watermark information to ensure that each local watermark information fragment can be recombined into the complete global watermark information according to the correct association relationship.

[0129] S505: Based on the correlation, the information content of multiple local watermark information is merged to restore the global watermark information.

[0130] Specifically, after establishing the association between local watermark information, the local information is combined and restored into complete global watermark information according to the association. First, the combination order or spatial arrangement of all local watermark information is obtained based on the association mapping table. When local watermark information is associated in a linear sequence, an empty global watermark information sequence is created, the length of which is equal to the sum of the lengths of all local watermark information. Following the order determined by the association, the content of the first local watermark information is written to the beginning of the global sequence, and subsequent local watermark information is sequentially appended to the global sequence. When local watermark information is associated in a two-dimensional matrix, a two-dimensional global watermark information matrix is ​​created. Based on the row and column index marked in the association, the content of each local watermark information is filled into the corresponding position in the global matrix. During the fusion process, the boundaries between local watermark information are processed, checking for redundant synchronization markers at the splicing points of adjacent local information and removing these redundant markers.

[0131] Based on the above embodiments, as an optional implementation method, the camera imaging parameters include the brand and model of the shooting device and the shooting condition parameters. In S106, the global watermark information is optimized according to the camera imaging parameters to obtain the target watermark information of the reproduced image. This can be achieved through the following steps S601-S603.

[0132] S601: Determine the imaging characteristics corresponding to the brand and model of the shooting equipment. The imaging characteristics include image compression characteristics and color deviation characteristics. Among them, the brand model refers to the combination of the manufacturer's name and the specific product model of the shooting equipment. Different brands and models of equipment have differences in image processing hardware and software algorithms.

[0133] Imaging characteristics refer to the inherent technical features and processing rules exhibited by a specific brand and model of shooting equipment in the process of converting optical signals into digital images. These characteristics are jointly determined by hardware and software factors such as the type of image sensor, image processing chip algorithm, and lens optical performance.

[0134] Image compression characteristics refer to the compression algorithm type, compression ratio, quantization step size, and other parameters used by the imaging device to compress and store the acquired raw image data. Differences in compression characteristics between different devices result in differences in the degree of image detail preservation and distortion mode.

[0135] Color deviation characteristics refer to the systematic deviation of shooting equipment from the colors of the real scene during the color reproduction process. These deviations include characteristics such as white balance algorithm tendencies, color saturation adjustment strategies, and color temperature compensation modes. These deviations exhibit specific regularities under different lighting conditions.

[0136] Specifically, after obtaining the brand and model information of the shooting equipment, it is necessary to determine the inherent imaging characteristics of the equipment that affect the watermark information. First, the records corresponding to the brand and model are retrieved from a pre-established equipment imaging characteristic database. This database was established through experimental testing and parameter calibration of various mainstream shooting equipment, recording the imaging characteristic parameters of different brands and models. Next, the image compression characteristic data of the equipment is read, including the default compression format, typical compression quality level, quantization table parameters, etc. These compression characteristics determine the equipment's ability to retain high-frequency detail information when storing images. Equipment with a high compression ratio causes greater loss of watermark information embedded in high-frequency components, while equipment with a low compression ratio retains watermark information more completely. Finally, the color deviation characteristic data of the equipment is read, including the color shift vector under standard light, the gain coefficient of each color channel, color space conversion matrix, etc. These color deviation characteristics determine the degree to which the equipment alters the watermark information embedded in color components. Equipment with large color deviation causes a large shift in the color encoding of the watermark information, while equipment with accurate color reproduction has less impact on the watermark color encoding.

[0137] S602: Based on the imaging characteristics and shooting condition parameters, determine the type and degree of influence of the shooting device on the global watermark information. The shooting condition parameters include at least one of the following: light intensity, shooting angle, and shooting distance. Illumination intensity refers to the energy density of light illuminating the surface of the subject in a shooting scene. Illumination intensity affects the overall brightness, contrast, and noise level of the image. Shooting angle refers to the angle between the optical axis of the shooting device and the normal direction of the plane being photographed. Shooting angle affects the degree of geometric distortion and perspective distortion in the image. Shooting distance refers to the spatial distance between the lens of the shooting device and the subject. Shooting distance affects the image's sharpness, depth of field, and detail resolution.

[0138] Specifically, after determining the imaging characteristics of the shooting equipment, the specific impact of the equipment on the global watermark information is analyzed in conjunction with the actual shooting conditions. First, the light intensity value in the shooting conditions is read, and the correlation analysis is performed between this light intensity and the color deviation characteristics in the equipment's imaging characteristics.

[0139] When the light intensity is in the low-light range, the device's automatic gain control is activated, leading to increased image noise. This noise interferes with the watermark signal, and the impact type is determined to be a decrease in the watermark signal-to-noise ratio. The noise amplitude is calculated based on the specific light intensity value and the device's noise characteristic curve, and the ratio of the noise amplitude to the watermark signal intensity is used as a quantitative indicator of the impact. When the light intensity deviates from the standard illuminance, the device's color compensation algorithm produces a color shift. This shift alters the watermark encoding embedded in the color channels, and the impact type is determined to be watermark color distortion. The color shift amount is calculated based on the degree of light intensity deviation and the color temperature compensation curve in the device's color deviation characteristics, and the shift amount is used as a quantitative indicator of the impact.

[0140] Read the shooting angle parameters. When the shooting angle deviates from the vertical direction, the image produces perspective distortion, which causes the spatial distribution of the watermark to be distorted. The type of influence is determined to be watermark geometric distortion. Based on the specific value of the shooting angle, the deformation coefficient of each region of the image is calculated through the perspective transformation model. The maximum value of the deformation coefficient is used as a quantitative indicator of the degree of influence.

[0141] The shooting distance parameter is read. When the shooting distance is too large, the image resolution is insufficient, resulting in the loss of watermark details. The impact type is determined to be watermark spectrum truncation. The effective frequency bandwidth of the image is calculated based on the shooting distance and the resolution characteristics of the device lens. The proportion of the watermark frequency band that exceeds the effective bandwidth is used as a quantitative indicator of the degree of impact.

[0142] S603: Adjust the extraction parameters of the global watermark information according to the type and degree of impact to obtain the target watermark information optimized for the shooting device.

[0143] Specifically, after clarifying the type and degree of impact of the shooting equipment on the global watermark information, the watermark extraction parameters are adjusted accordingly to compensate for these impacts. First, addressing the type of impact of reduced watermark signal-to-noise ratio, the parameter settings of the noise reduction filter are adjusted. Based on the noise amplitude value in the impact severity index, the noise reduction intensity of the filter is increased. This is achieved by increasing the filter's cutoff frequency or increasing the number of filters to suppress noise interference with the watermark signal. Simultaneously, the signal detection threshold during watermark decoding is lowered. The threshold value is reduced proportionally to the degree of signal-to-noise ratio reduction, ensuring that even weak watermark signals can still be effectively detected.

[0144] To address the types of color distortion affecting watermarks, the color compensation coefficient is adjusted. Based on the color offset in the impact index, a reverse compensation coefficient is calculated. Before extracting the watermark, a reverse compensation transformation is applied to the color components of the global watermark information to correct the offset color values ​​back to their original state.

[0145] To address the impact of watermark geometric distortion, the parameters of the geometric correction transformation are adjusted. Based on the deformation coefficient in the impact degree index, the inverse perspective transformation matrix is ​​calculated to perform geometric correction processing on the global watermark information, restoring the distorted watermark spatial distribution to a normal form.

[0146] To address the impact types of watermark spectral truncation, frequency domain reconstruction parameters are adjusted. Based on the proportion of spectral loss in the impact index, either a spectral extrapolation algorithm is enabled or the dependence on high-frequency components is reduced. Watermark information is extracted from the retained spectral components by increasing the weight of low-frequency components or using partial frequency band decoding methods. When multiple impact types exist, the corresponding extraction parameters are adjusted sequentially in descending order of impact severity, prioritizing compensation for the most severe distortion type.

[0147] The following are system embodiments of this application, which can be used to execute the method embodiments of this application. For details not disclosed in the system embodiments of this application, please refer to the method embodiments of the application.

[0148] Please see Figure 2 This illustration shows a schematic diagram of a screen-hidden watermark extraction and tracing system provided in an exemplary embodiment of this application. The system can be implemented through software, hardware, or a combination of both, forming all or part of a larger system. The screen-hidden watermark extraction and tracing system includes: The image input module is used to acquire a copy of the image generated after the screen is captured by a camera. The grid division module is used to obtain screen parameters from the reproduced image and divide the reproduced image into multiple grid units of equal size according to the screen parameters; The confidence calculation module is used to perform feature analysis on the image of each grid cell and output the watermark recognition confidence score corresponding to each grid cell. The grid filtering module is used to identify grid cells with a watermark recognition confidence level greater than a preset confidence threshold as target grid cells, and to match the feature patterns of the target grid cells with a preset watermark template to determine the local watermark information of the target grid cells. The fusion module is used to fuse multiple local watermark information based on the splicing position obtained after merging and splicing multiple adjacent target grid cells to obtain global watermark information. The optimization processing module is used to obtain the camera imaging parameters of the shooting device, and optimize the global watermark information according to the camera imaging parameters to obtain the target watermark information of the reproduced image. The target watermark information is used to trace the reproduction information of the reproduced image.

[0149] This application also provides a computer storage medium that can store multiple instructions. The instructions are adapted to be loaded by a processor and executed as described in the above embodiments for the method of extracting and tracing hidden watermarks on the screen. For details of the execution process, please refer to the specific description of the embodiments, which will not be repeated here.

[0150] Please see Figure 3 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Figure 3 As shown, the electronic device 300 may include: at least one processor 301, at least one network interface 304, user interface 303, memory 305, and at least one communication bus 302.

[0151] The communication bus 302 is used to enable communication between these components.

[0152] The user interface 303 may include a display screen and a camera.

[0153] The network interface 304 may optionally include a standard wired interface or a wireless interface (such as a Wi-Fi interface).

[0154] The processor 301 may include one or more processing cores. The processor 301 connects to various parts of the server using various interfaces and lines, and performs various server functions and processes data by running or executing instructions, programs, code sets, or instruction sets stored in the memory 305, and by calling data stored in the memory 305. Optionally, the processor 301 may be implemented using at least one hardware form of digital signal processing, field-programmable gate array, or programmable logic array. The processor 301 may integrate one or more of the following: a central processing unit (CPU), a graphics processing unit (GPU), and a modem. The CPU primarily handles the operating system, user interface, and applications; the GPU is responsible for rendering and drawing the content required for display; and the modem handles wireless communication. It is understood that the modem may also not be integrated into the processor 301 and may be implemented as a separate chip.

[0155] The memory 305 may include random access memory (RAM) or read-only memory (ROM). Optionally, the memory 305 may include a non-transitory computer-readable medium. The memory 305 may be used to store instructions, programs, code, code sets, or instruction sets. The memory 305 may include a program storage area and a data storage area. The program storage area may store instructions for implementing an operating system, instructions for at least one function (such as touch functionality, sound playback functionality, image playback functionality, etc.), instructions for implementing the various method embodiments described above, etc.; the data storage area may store data involved in the various method embodiments described above, etc. Optionally, the memory 305 may also be at least one storage device located remotely from the aforementioned processor 301. Figure 3 As shown, the memory 305, which serves as a computer storage medium, may include an operating system, a network communication module, a user interface module, and an application program for extracting and tracing hidden watermarks on a screen.

[0156] exist Figure 3 In the electronic device 300 shown, the user interface 303 is mainly used to provide an input interface for the user and to obtain the user input data; while the processor 301 can be used to call an application program stored in the memory 305 that is a method for extracting and tracing a hidden screen watermark. When executed by one or more processors, the electronic device executes one or more methods as described in the above embodiments.

[0157] An electronic device readable storage medium stores instructions that, when executed by one or more processors, cause the electronic device to perform one or more methods as described in the above embodiments.

[0158] It should be noted that, for the sake of simplicity, the foregoing method embodiments are all described as a series of actions. However, those skilled in the art should understand that this application is not limited to the described order of actions, as some steps may be performed in other orders or simultaneously according to this application. Furthermore, those skilled in the art should also understand that the embodiments described in the specification are preferred embodiments, and the actions and modules involved are not necessarily essential to this application.

[0159] In the above embodiments, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions in other embodiments.

[0160] In the several embodiments provided in this application, it should be understood that the disclosed apparatus can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the shown or discussed mutual couplings or direct couplings or communication connections may be through some service interfaces; indirect couplings or communication connections between apparatuses or units may be electrical or other forms.

[0161] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0162] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0163] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage device (CMD). Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a memory and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of this application. The aforementioned memory includes various media capable of storing program code, such as USB flash drives, portable hard drives, magnetic disks, or optical disks.

[0164] The above are merely exemplary embodiments of this disclosure and should not be construed as limiting the scope of this disclosure. Any equivalent changes and modifications made in accordance with the teachings of this disclosure shall still fall within the scope of this disclosure. Those skilled in the art will readily conceive of other embodiments of this disclosure upon considering the specification and the disclosure of practical truths. This application is intended to cover any variations, uses, or adaptations of this disclosure that follow the general principles of this disclosure and include common knowledge or customary techniques in the art not described in this disclosure.

Claims

1. A method for extracting and tracing the source of hidden watermarks on a screen, characterized in that, The method includes: To obtain a reproduced image of the screen after it has been photographed using a camera; Obtain the screen parameters in the reproduced image, and divide the reproduced image into multiple grid units of equal size according to the screen parameters; Perform feature analysis on the image of each grid cell and output the watermark recognition confidence level corresponding to each grid cell; The step of performing feature analysis on the image of each grid cell and outputting the watermark recognition confidence score corresponding to each grid cell includes: Feature analysis is performed on the image of each grid cell to extract image features of each grid cell; watermark quality features are identified in the image features of each grid cell to determine the initial confidence value of each grid cell; for each grid cell, watermark candidate features of the grid cells spatially adjacent to the grid cell are obtained, and the similarity between the watermark candidate features of the grid cell and the watermark candidate features of the adjacent grid cells is calculated; the initial confidence value is adjusted according to the similarity to obtain the watermark recognition confidence value corresponding to each grid cell; The grid cells with a watermark recognition confidence level greater than a preset confidence threshold are identified as target grid cells, and the feature patterns of the target grid cells are matched with a preset watermark template to determine the local watermark information of the target grid cells. Based on the splicing position obtained after merging and splicing multiple adjacent target grid cells, the multiple local watermark information is fused to obtain global watermark information; The camera imaging parameters of the shooting device are obtained, and the global watermark information is optimized based on the camera imaging parameters to obtain the target watermark information of the reproduced image. The target watermark information is used to trace the reproduction information of the reproduced image.

2. The method according to claim 1, characterized in that, The step of dividing the reproduced image into multiple equally sized grid units according to the screen parameters includes: Identify the screen boundaries in the reproduced image to determine the effective display area of ​​the screen; Based on the screen parameters, the corresponding screen model is matched from the preset screen parameter database to determine the display characteristic parameters corresponding to the screen model; The grid cell size that matches the screen resolution is determined based on the display characteristic parameters, and the effective display area is uniformly divided according to the grid cell size to obtain multiple grid cells of equal size.

3. The method according to claim 2, characterized in that, The method further includes: Collect parameter information of screens of various brands, models or specifications, including at least one of screen resolution, color space type, brightness range, contrast range, refresh rate and screen display technology type; Based on the parameter information of each screen, the display characteristic simulation parameters of each screen are determined, and the display influence coefficient of each screen on the embedded watermark is calculated based on the display characteristic simulation parameters. The display influence coefficient represents the degree of difference between the image before and after the watermark is embedded. Based on the numerical range of the resolution of each screen, the corresponding screens are divided into multiple resolution levels, and each resolution level is set with a corresponding grid cell size parameter; Establish and store the corresponding relationships between the parameter information of each screen, the display influence coefficient, and the grid unit size parameter to construct a screen parameter database.

4. The method according to claim 1, characterized in that, The step of identifying watermark quality features in the image features of each grid cell and determining the initial confidence value of each grid cell includes: The image features of each grid cell are compared with a preset watermark quality feature template, and the feature parts of the image features of each grid cell that conform to the preset watermark quality feature template are identified as the watermark quality features of each grid cell. The initial confidence value of each grid cell is calculated based on the degree of matching between the watermark quality features of each grid cell and the preset watermark feature template.

5. The method according to claim 1, characterized in that, The process involves fusing multiple local watermark information entries based on the splicing positions obtained after merging and splicing multiple adjacent target grid cells to obtain global watermark information, including: Acquire spatial coordinate information of multiple adjacent target grid units in the photographed image, and determine the relative positional relationship between each target grid unit; Based on the relative positional relationship, multiple adjacent target grid cells are merged and spliced ​​to obtain the splicing position of the target grid cells in the spliced ​​target area; Obtain the information content of the local watermark information corresponding to each target grid cell; The correlation between the local watermark information is determined based on the splicing position of each target grid cell; Based on the aforementioned association, the information content of multiple local watermark information is fused to restore the global watermark information.

6. The method according to claim 1, characterized in that, The camera imaging parameters include the brand and model of the shooting device and shooting condition parameters. The step of optimizing the global watermark information based on the camera imaging parameters to obtain the target watermark information of the reproduced image includes: Based on the brand and model of the shooting device, determine the imaging characteristics corresponding to the brand and model, including image compression characteristics and color deviation characteristics; Based on the imaging characteristics and the shooting condition parameters, determine the type and degree of influence of the shooting device on the global watermark information. The shooting condition parameters include at least one of light intensity, shooting angle and shooting distance. Based on the type and degree of influence, the extraction parameters of the global watermark information are adjusted to obtain target watermark information optimized for the shooting device.

7. A system for extracting and tracing hidden watermarks on a screen, characterized in that, The system includes: The image input module is used to acquire a copy of the image generated after the screen is captured by a camera. The grid division module is used to obtain the screen parameters in the reproduced image and divide the reproduced image into multiple grid units of equal size according to the screen parameters; The confidence calculation module is used to perform feature analysis on the image of each grid cell and output the watermark recognition confidence score corresponding to each grid cell. The feature analysis on the image of each grid cell and the output of the watermark recognition confidence score corresponding to each grid cell includes: performing feature analysis on the image of each grid cell to extract image features of each grid cell; identifying watermark quality features in the image features of each grid cell to determine the initial confidence score value of each grid cell; for each grid cell, obtaining watermark candidate features of grid cells spatially adjacent to the grid cell, calculating the similarity between the watermark candidate features of the grid cell and the watermark candidate features of the adjacent grid cells; and adjusting the initial confidence score based on the similarity to obtain the watermark recognition confidence score corresponding to each grid cell. The grid filtering module is used to identify grid cells with a watermark recognition confidence level greater than a preset confidence threshold as target grid cells, and to match the feature patterns of the target grid cells with a preset watermark template to determine the local watermark information of the target grid cells. The fusion module is used to fuse multiple local watermark information based on the splicing position obtained after merging and splicing multiple adjacent target grid cells to obtain global watermark information. An optimization processing module is used to acquire the camera imaging parameters of the shooting device, and optimize the global watermark information according to the camera imaging parameters to obtain the target watermark information of the reproduced image. The target watermark information is used to trace the reproduction information of the reproduced image.

8. A computer storage medium, characterized in that, The computer storage medium stores a plurality of instructions, which are adapted to be loaded by a processor and executed as described in any one of claims 1 to 6.

9. An electronic device, characterized in that, The device includes a processor, a memory, and a transceiver, wherein the memory is used to store instructions, the transceiver is used to communicate with other devices, and the processor is used to execute the instructions stored in the memory to cause the electronic device to perform the method as described in any one of claims 1 to 6.

Images