A machine learning-based image data processing system and method

Abstract

The application relates to the technical field of data processing, and discloses a picture-text-video data processing system and method based on machine learning, which comprises a collection layer configured to analyze image feature parameters, determine an initial compression ratio of to-be-processed picture-text-video data based on an analysis result, a judgment layer configured to collect regional feature information of each monitoring area and determine whether the initial compression ratio is optimized according to the regional feature information, an optimization layer configured to determine an optimization coefficient of the initial compression ratio according to the regional feature information and real-time network data, and obtain an optimized compression ratio, and a display layer configured to compress the to-be-processed picture-text-video data according to the optimized compression ratio, and transmit the compressed to-be-processed picture-text-video data. The application effectively improves the processing efficiency and compression quality of picture-text-video data.

Description

Technical Field

[0001] The present invention relates to the field of data processing technology, and in particular to a system and method for processing graphic and text image data based on machine learning. Background Art

[0002] During the transmission of graphic, text, and image data, compression is often required to reduce data size and improve transmission efficiency. However, while traditional compression technologies reduce data size, they often result in the loss of details in the graphic, text, and image data, which undoubtedly has a negative impact on the overall data quality.

[0003] Therefore, it is necessary to design a graphic image data processing system and method based on machine learning to solve the problems existing in current technology. Summary of the Invention

[0004] In view of this, the present invention proposes a system and method for processing graphic and text image data based on machine learning, aiming to perform intelligent processing of graphic and text image data through machine learning algorithms, thereby achieving efficient compression while maintaining high data quality.

[0005] In one aspect, the present invention provides a system for processing graphic and text image data based on machine learning, comprising:

[0006] an acquisition layer configured to determine image data to be processed, acquire image feature parameters of the image data to be processed, analyze the image feature parameters, and determine an initial compression ratio of the image data to be processed based on the analysis results; wherein the initial compression ratio is a ratio between high-quality compression and low-quality compression of the image data to be processed;

[0007] The judgment layer is configured to divide the image data to be processed into regions based on an image recognition model to obtain a plurality of monitoring regions; collect regional feature information of each monitoring region, and determine whether to optimize the initial compression ratio based on the regional feature information;

[0008] an optimization layer configured to, when determining to optimize the initial compression ratio, collect real-time network data, determine an optimization coefficient of the initial compression ratio based on the regional feature information and the real-time network data, and obtain an optimized compression ratio;

[0009] The presentation layer is configured to compress the image data to be processed according to the optimized compression ratio and transmit the compressed image data to be processed.

[0010] Furthermore, the acquisition layer analyzes the image feature parameters and determines the initial compression ratio of the to-be-processed graphic image data based on the analysis result, including:

[0011] Analyzing the image characteristic parameters to obtain the image type, transmission priority, image resolution, data volume, image noise level value, and color depth characteristic value of the image data to be processed;

[0012] determining a compression type according to the image type;

[0013] Constructing the compression type, transmission priority, image resolution, data volume, image noise level value and color depth feature value into an initial compression vector group;

[0014] The initial compression vector group is compared with historical data, and an initial compression ratio of the to-be-processed graphic image data is determined according to the comparison result.

[0015] Furthermore, when the acquisition layer determines the initial compression ratio of the to-be-processed graphic image data according to the comparison result, it includes:

[0016] When there exists a historical compression vector group identical to the initial compression vector group in the historical data, using a historical compression ratio corresponding to the historical compression vector group as the initial compression ratio;

[0017] When there is no historical compression vector group identical to the initial compression vector group in the historical data, historical compression vector groups corresponding to the same compression type and transmission priority are screened out and constructed as a comparison vector group; and matching degrees between the initial compression vector group and the comparison vector group are calculated one by one, and the initial compression ratio is determined based on the matching degrees.

[0018] Furthermore, when the acquisition layer determines the initial compression ratio according to the matching degree, it includes:

[0019] Extract the maximum value of the matching degree and record it as the maximum matching degree;

[0020] Comparing the maximum matching degree with a first maximum matching degree and a second maximum matching degree, and determining the initial compression ratio according to the comparison result; wherein the first maximum matching degree is less than the second maximum matching degree;

[0021] When the maximum matching degree is less than or equal to the first maximum matching degree, determining the initial compression ratio to be the first compression ratio;

[0022] When the maximum matching degree is greater than the first maximum matching degree and less than or equal to the second maximum matching degree, determining the initial compression ratio to be a second compression ratio; wherein the second compression ratio is less than the first compression ratio;

[0023] When the maximum matching degree is greater than the second maximum matching degree, the initial compression ratio is determined to be a third compression ratio; wherein the third compression ratio is less than the second compression ratio.

[0024] Furthermore, the judgment layer divides the image data to be processed into regions based on the image recognition model to obtain a number of monitoring regions, including:

[0025] Using the image recognition model to divide the image data to be processed into regions to obtain a number of monitoring areas;

[0026] Feature extraction is performed on each of the monitoring areas to obtain regional feature information corresponding to each of the monitoring areas.

[0027] Furthermore, when the judgment layer judges whether to optimize the initial compression ratio according to the regional feature information, it includes:

[0028] Collecting regional characteristic information corresponding to each of the monitoring areas, wherein the regional characteristic information includes text density, image complexity, color richness, and key information distribution rate;

[0029] Obtaining characteristic standard values ​​corresponding to the regional characteristic information, the characteristic standard values ​​including standard text density, standard image complexity, standard color richness, and standard key information distribution rate;

[0030] Counting the number of regional feature information in each monitoring area that is greater than or equal to the corresponding feature standard value, and recording it as the feature number;

[0031] Determine the level of the current monitoring area according to the number of features;

[0032] If the number of features is less than 3, the level of the monitoring area is determined to be the first level;

[0033] If the number of features is equal to 3, determining that the level of the monitoring area is the second level;

[0034] If the number of features is equal to 4, determining that the level of the monitoring area is the third level;

[0035] The number and value of the monitoring areas corresponding to the second level and the third level are counted, and whether to optimize the initial compression ratio is determined according to the number and value.

[0036] Furthermore, when the judgment layer judges whether to optimize the initial compression ratio according to the quantity and value, it includes:

[0037] Comparing the quantity and value with a quantity threshold, and determining whether to optimize the initial compression ratio based on the comparison result;

[0038] When the sum of the quantity and the value is greater than the quantity threshold, determining to optimize the initial compression ratio;

[0039] Otherwise, it is determined that the initial compression ratio is not to be optimized, and the initial compression ratio is used as the optimized compression ratio.

[0040] Furthermore, the optimization layer determines the optimization coefficient of the initial compression ratio according to the regional feature information and the real-time network data, and obtains the optimized compression ratio, including:

[0041] Calculating the difference between each regional characteristic information and the corresponding characteristic standard value in each of the monitoring areas corresponding to the second level and the third level, and constructing all the differences into a first set;

[0042] Extracting features from the real-time network data to obtain a number of network feature values;

[0043] Get the network standard value corresponding to each network characteristic value;

[0044] Calculate the difference between each network characteristic value and the corresponding network standard value, and construct all the differences into a second set;

[0045] Calculating a feature offset value according to the first set, calculating a network offset value according to the second set, and calculating a comprehensive offset value according to the feature offset value and the network offset value;

[0046] determining an optimization coefficient of the initial compression ratio based on the comprehensive offset value;

[0047] The product value of the optimization coefficient and the initial compression ratio is used as the optimized compression ratio.

[0048] Furthermore, when the optimization layer determines the optimization coefficient of the initial compression ratio based on the comprehensive offset value, it includes:

[0049] Comparing the comprehensive offset value with a first comprehensive offset value and a second comprehensive offset value, and determining an optimization coefficient of the initial compression ratio according to the comparison result; wherein the first comprehensive offset value is smaller than the second comprehensive offset value;

[0050] When the comprehensive offset value is less than or equal to the first comprehensive offset value, determining the optimization coefficient of the initial compression ratio to be the first optimization coefficient;

[0051] When the comprehensive offset value is greater than the first comprehensive offset value and less than or equal to the second comprehensive offset value, determining the optimization coefficient of the initial compression ratio to be a second optimization coefficient; wherein the second optimization coefficient is less than the first optimization coefficient;

[0052] When the comprehensive offset value is greater than the second comprehensive offset value, the optimization coefficient of the initial compression ratio is determined to be a third optimization coefficient; wherein the third optimization coefficient is smaller than the second optimization coefficient.

[0053] Compared with existing technologies, the present invention offers the following advantages: the machine learning-based image data processing system provided by the present invention effectively improves the processing efficiency and compression quality of image data through an intelligent processing process. Specifically, the system automatically determines the appropriate initial compression ratio for different types of image data and, through region segmentation and feature extraction, accurately determines whether further optimization of the compression ratio is necessary. Furthermore, with the assistance of real-time network data, the system can dynamically adjust the optimization coefficient to ensure that the resulting optimized compression ratio meets the requirements for efficient compression while maintaining high data quality.

[0054] In another aspect, the present invention also proposes a method for processing graphic image data based on machine learning, comprising the following steps:

[0055] Determining image data to be processed, collecting image feature parameters of the image data to be processed, analyzing the image feature parameters, and determining an initial compression ratio of the image data to be processed based on the analysis results; wherein the initial compression ratio is a ratio of high-quality compression to low-quality compression of the image data to be processed;

[0056] Dividing the image data to be processed into regions based on an image recognition model to obtain a plurality of monitoring regions; collecting regional feature information of each monitoring region, and determining whether to optimize the initial compression ratio based on the regional feature information;

[0057] When it is determined that the initial compression ratio is to be optimized, real-time network data is collected, and an optimization coefficient of the initial compression ratio is determined based on the regional characteristic information and the real-time network data, and an optimized compression ratio is obtained;

[0058] The image data to be processed is compressed according to the optimized compression ratio, and the compressed image data to be processed is transmitted.

[0059] It is understandable that the above-mentioned machine learning-based graphic image data processing system and method have the same beneficial effects and will not be repeated here. BRIEF DESCRIPTION OF THE DRAWINGS

[0060] Various other advantages and benefits will become apparent to those skilled in the art upon reading the detailed description of the preferred embodiment below. The accompanying drawings are for illustration purposes only and are not to be considered as limiting the present invention. The same reference symbols are used throughout the drawings to represent the same components. In the drawings:

[0061] Figure 1 A structural block diagram of a machine learning-based graphic and image data processing system provided by an embodiment of the present invention;

[0062] Figure 2 A flowchart of a method for processing graphic and text image data based on machine learning provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0063] Exemplary embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although exemplary embodiments of the present disclosure are shown in the accompanying drawings, it should be understood that the present disclosure can be implemented in various forms and should not be limited by the embodiments set forth herein. On the contrary, these embodiments are provided to enable a more thorough understanding of the present disclosure and to fully convey the scope of the present disclosure to those skilled in the art. It should be noted that, unless there is a conflict, the embodiments of the present disclosure and the features described in the embodiments may be combined with each other. The present invention will be described in detail below with reference to the accompanying drawings and in conjunction with the embodiments.

[0064] See Figure 1 As shown, in some embodiments of the present application, this embodiment provides a graphic image data processing system based on machine learning, including:

[0065] The acquisition layer is configured to determine the image data to be processed, collect image feature parameters of the image data to be processed, analyze the image feature parameters, and determine an initial compression ratio of the image data to be processed based on the analysis results; wherein the initial compression ratio is a ratio of high-quality compression to low-quality compression of the image data to be processed;

[0066] The judgment layer is configured to divide the image data to be processed into regions based on the image recognition model to obtain a number of monitoring regions; collect regional feature information of each monitoring region, and determine whether to optimize the initial compression ratio based on the regional feature information;

[0067] The optimization layer is configured to, when determining to optimize the initial compression ratio, collect real-time network data, determine an optimization coefficient of the initial compression ratio based on the regional feature information and the real-time network data, and obtain an optimized compression ratio;

[0068] The presentation layer is configured to compress the image data to be processed according to an optimized compression ratio and transmit the compressed image data to be processed.

[0069] It is understood that the machine learning-based image data processing system provided in this embodiment effectively improves the processing efficiency and compression quality of image data through an intelligent processing process. Specifically, the system can automatically determine the appropriate initial compression ratio for different types of image data and, through region segmentation and feature extraction, accurately determine whether the compression ratio needs to be further optimized. Furthermore, with the assistance of real-time network data, the system can dynamically adjust the optimization coefficient to ensure that the final optimized compression ratio meets the requirements for efficient compression while maintaining high data quality.

[0070] Specifically, the acquisition layer analyzes the image feature parameters and determines the initial compression ratio of the image data to be processed based on the analysis results, including:

[0071] Analyze the image characteristic parameters to obtain the image type, transmission priority, image resolution, data volume, image noise level value and color depth characteristic value of the image data to be processed;

[0072] Determine the compression type based on the image type;

[0073] Constructing compression type, transmission priority, image resolution, data volume, image noise level value and color depth feature value into an initial compression vector group;

[0074] The initial compression vector group is compared with the historical data, and the initial compression ratio of the image data to be processed is determined according to the comparison result.

[0075] In this embodiment, image types include text images, chart images, photo images, and hand-drawn images. When the image type is a text image, the compression type is determined to be lossless compression; when the image type is a chart image, photo image, or hand-drawn image, the compression type is determined to be lossy compression.

[0076] In this embodiment, the image noise level value is calculated by an image noise detection algorithm, specifically, the image data to be processed is preprocessed to remove noise points therein, and then the variance of the remaining noise is calculated using the image noise detection algorithm, and the variance is used as the image noise level value.

[0077] In this embodiment, the color depth characteristic value is preferably the color bit number (such as 8 bits, 24 bits, etc.) of the graphic image data to be processed, which is used to represent the number of bits occupied by the color information contained in each pixel in the image.

[0078] Specifically, when the acquisition layer determines the initial compression ratio of the image data to be processed based on the comparison results, it includes:

[0079] When there is a historical compression vector group that is the same as the initial compression vector group in the historical data, the historical compression ratio corresponding to the historical compression vector group is used as the initial compression ratio;

[0080] When there is no historical compression vector group identical to the initial compression vector group in the historical data, historical compression vector groups corresponding to the same compression type and transmission priority are screened out and constructed as a comparison vector group; the matching degree between the initial compression vector group and the comparison vector group is calculated one by one, and the initial compression ratio is determined based on the matching degree.

[0081] In this embodiment, the matching degree is obtained by the following formula:

[0082] ;

[0083] Among them, Pi represents the matching degree between the initial vector group and the i-th comparison vector group; xj represents the j-th image feature parameter of the initial compressed vector group; yij represents the j-th image feature parameter of the i-th comparison vector group; ωj represents the weight of the j-th image feature parameter; n represents the number of image feature parameters.

[0084] As you can see, by calculating the matching degree, the system can accurately assess the similarity between the initial compression vector group and each comparison vector group in the historical data, thereby reasonably determining the initial compression ratio. This intelligent comparison method based on historical data not only improves the efficiency of compression ratio determination, but also ensures its accuracy and applicability.

[0085] As you can see, by comparing the initial compression vector set with historical compression vector sets, the system can quickly identify historical processing cases similar to the image data being processed, accurately determining an initial compression ratio that is both efficient and ensures quality. This approach not only significantly reduces the need for manual intervention, but also significantly improves processing efficiency and accuracy.

[0086] Specifically, the collection layer determines the initial compression ratio based on the matching degree, including:

[0087] Extract the maximum value of the matching degree and record it as the maximum matching degree;

[0088] Comparing the maximum matching degree with the first maximum matching degree and the second maximum matching degree, and determining an initial compression ratio according to the comparison result; wherein the first maximum matching degree is less than the second maximum matching degree;

[0089] When the maximum matching degree is less than or equal to the first maximum matching degree, determining the initial compression ratio to be the first compression ratio;

[0090] When the maximum matching degree is greater than the first maximum matching degree and less than or equal to the second maximum matching degree, determining the initial compression ratio to be the second compression ratio; wherein the second compression ratio is less than the first compression ratio;

[0091] When the maximum matching degree is greater than the second maximum matching degree, the initial compression ratio is determined to be a third compression ratio; wherein the third compression ratio is less than the second compression ratio.

[0092] In this embodiment, the preferred value of the first maximum matching degree is 0.85, and the preferred value of the second maximum matching degree is 0.95.

[0093] It can be understood that by setting different matching thresholds (the first maximum matching threshold and the second maximum matching threshold), the system can flexibly adjust the initial compression ratio determination strategy to accommodate image data of varying complexity and quality requirements. When the maximum matching threshold is low, the system tends to select a more relaxed compression ratio (i.e., the first compression ratio) to ensure data integrity and readability. When the maximum matching threshold is between the first and second maximum matching thresholds, the system selects a moderate compression ratio (i.e., the second compression ratio) to strike a balance between data quality and compression efficiency. When the maximum matching threshold is high, the system selects a stricter compression ratio (i.e., the third compression ratio) to maximize compression efficiency while maintaining basic data quality.

[0094] Specifically, the judgment layer divides the image data to be processed into regions based on the image recognition model, and obtains several monitoring areas, including:

[0095] Use the image recognition model to divide the image data to be processed into regions and obtain several monitoring areas;

[0096] Feature extraction is performed on each monitoring area to obtain regional feature information corresponding to each monitoring area.

[0097] In this embodiment, the image recognition model is preferably a deep learning model, such as a convolutional neural network (CNN) or a residual network (ResNet). These models are trained with a large amount of graphic and image data and can accurately identify different contents and features in the image, thereby achieving accurate area division.

[0098] It's understandable that the judgment layer, through intelligent region division and feature extraction, can accurately identify key regions and features within image data, providing strong support for subsequent optimization processing. This region division approach not only improves the targeted nature of processing but also effectively avoids unnecessary data loss. Furthermore, when collecting regional feature information for each monitored area, the judgment layer fully considers the details and differences in the image, ensuring the accuracy and completeness of this regional feature information.

[0099] Specifically, the judgment layer determines whether to optimize the initial compression ratio based on regional feature information, including:

[0100] Collect regional characteristic information corresponding to each monitoring area, including text density, image complexity, color richness and key information distribution rate;

[0101] Obtaining characteristic standard values ​​corresponding to regional characteristic information, the characteristic standard values ​​including standard text density, standard image complexity, standard color richness, and standard key information distribution rate;

[0102] Count the number of regional feature information in each monitoring area that is greater than or equal to the corresponding feature standard value and record it as the feature number;

[0103] Determine the level of the current monitoring area based on the number of features;

[0104] If the number of features is less than 3, the level of the monitoring area is determined to be the first level;

[0105] If the number of features is equal to 3, the level of the monitoring area is determined to be the second level;

[0106] If the number of features is equal to 4, the level of the monitoring area is determined to be the third level;

[0107] The number and value of the monitoring areas corresponding to the second level and the third level are counted, and whether the initial compression ratio should be optimized is determined based on the number and value.

[0108] In this embodiment, the image complexity is preferably image entropy.

[0109] In this embodiment, color richness refers to the diversity and range of colors in an image. Specifically, color richness can be evaluated by calculating the frequency and distribution of different colors in the image.

[0110] Color richness can be obtained by the following formula:

[0111] ;

[0112] Among them, C represents the color richness; qk represents the frequency of the kth color in the image; N represents the total number of colors in the image.

[0113] In this embodiment, key information includes characters, text, symbols, logos, charts, and other elements that are important for image understanding and recognition. The key information distribution rate refers to the area ratio occupied by key information in the image.

[0114] It is understandable that the judgment layer conducts a comprehensive and detailed analysis of the characteristics of the graphic image data by comprehensively considering multiple dimensions such as text density, image complexity, color richness, and key information distribution rate. This multi-dimensional analysis method not only improves the accuracy of judgment, but also effectively avoids misjudgment or missed judgment due to single-dimensional judgment. When the amount of regional feature information in the monitoring area is large, that is, the image complexity is high, the color richness is large, or the key information distribution is denser, the system believes that these areas may contain more important information, so it is necessary to optimize the initial compression ratio to ensure that this information can be better retained during the compression process. On the contrary, if the amount of regional feature information in the monitoring area is small, the system may believe that the information importance of these areas is relatively low, and the compression ratio can be appropriately relaxed to reduce the amount of data and improve processing efficiency.

[0115] Specifically, the judgment layer determines whether to optimize the initial compression ratio based on the quantity and value, including:

[0116] Compare the quantity and value with the quantity threshold, and determine whether to optimize the initial compression ratio based on the comparison result;

[0117] When the quantity and value are greater than the quantity threshold, it is determined that the initial compression ratio is optimized;

[0118] Otherwise, it is determined that the initial compression ratio is not to be optimized, and the initial compression ratio is used as the optimized compression ratio.

[0119] Specifically, the optimization layer determines the optimization coefficient of the initial compression ratio based on regional feature information and real-time network data, and obtains the optimized compression ratio, including:

[0120] Calculate the difference between the characteristic information of each area and the corresponding characteristic standard value in each monitoring area corresponding to the second level and the third level, and construct all the differences into a first set;

[0121] Extract features from real-time network data to obtain several network feature values;

[0122] Get the network standard value corresponding to each network characteristic value;

[0123] Calculate the difference between each network characteristic value and the corresponding network standard value, and construct all the differences into a second set;

[0124] Calculating a feature offset value according to the first set, calculating a network offset value according to the second set, and calculating a comprehensive offset value according to the feature offset value and the network offset value;

[0125] determining an optimization coefficient of an initial compression ratio based on the comprehensive offset value;

[0126] The product of the optimization coefficient and the initial compression ratio is taken as the optimized compression ratio.

[0127] In this embodiment, the characteristic offset value is obtained by the following formula:

[0128] ;

[0129] Wherein, Df represents the characteristic offset value; fbr represents the characteristic information of the rth region in the bth monitoring area; sr represents the characteristic standard value corresponding to the characteristic information of the rth region in the bth monitoring area; K represents the number and value of the monitoring areas corresponding to the second and third levels; M represents that there are M regional characteristic information in the bth monitoring area.

[0130] In this embodiment, the network offset value is obtained by the following formula:

[0131] ;

[0132] Where Dw represents the network offset value; wτ represents the τth network eigenvalue; tτ represents the network standard value corresponding to the τth network eigenvalue; and L represents the number of network eigenvalues.

[0133] In this embodiment, the comprehensive offset value is obtained by the following formula:

[0134] ;

[0135] Where Dc represents the comprehensive offset value; α and β represent the weighted exponents of the feature offset value and the network offset value, respectively, and α+β=1.

[0136] It's understandable that the optimization layer fully considers the characteristic offset of the image data and the network offset of the real-time network data when determining the optimization coefficient. The characteristic offset reflects the degree of difference between the image data itself and the standard characteristics, while the network offset reflects the potential demand and trend of data compression in the current network environment. By combining these two factors, the system can more comprehensively assess the adaptability of the initial compression ratio and make more appropriate optimization adjustments accordingly. This approach not only improves the accuracy and pertinence of the compression ratio, but also effectively enhances the system's adaptability to complex and changing environments.

[0137] Specifically, when the optimization layer determines the optimization coefficient of the initial compression ratio based on the comprehensive offset value, it includes:

[0138] Comparing the comprehensive offset value with the first comprehensive offset value and the second comprehensive offset value, and determining an optimization coefficient of the initial compression ratio according to the comparison result; wherein the first comprehensive offset value is smaller than the second comprehensive offset value;

[0139] When the comprehensive offset value is less than or equal to the first comprehensive offset value, determining the optimization coefficient of the initial compression ratio as the first optimization coefficient;

[0140] When the comprehensive offset value is greater than the first comprehensive offset value and less than or equal to the second comprehensive offset value, determining the optimization coefficient of the initial compression ratio to be the second optimization coefficient; wherein the second optimization coefficient is less than the first optimization coefficient;

[0141] When the comprehensive offset value is greater than the second comprehensive offset value, the optimization coefficient of the initial compression ratio is determined to be the third optimization coefficient; wherein the third optimization coefficient is smaller than the second optimization coefficient.

[0142] As can be understood, by setting different comprehensive offset thresholds, the optimization layer can flexibly adjust the optimization coefficient based on different situations, achieving more precise compression ratio optimization. When the comprehensive offset value is small, it indicates that the characteristics of the image data are only slightly different from the standard characteristics and real-time network data. In this case, the system tends to select a larger optimization coefficient (i.e., the first optimization coefficient) to preserve the original data quality and detail. When the comprehensive offset value is moderate, it indicates that the characteristics of the image data are moderately different from real-time network data. In this case, the system selects an intermediate optimization coefficient (i.e., the second optimization coefficient) to achieve a balance between data compression and transmission. When the comprehensive offset value is large, it indicates that the characteristics of the image data are significantly different from the standard characteristics and real-time network data. In this case, the system selects a smaller optimization coefficient (i.e., the third optimization coefficient) to better consider data compression requirements, reduce data volume, and improve transmission efficiency. This method of flexibly adjusting the optimization coefficient based on the comprehensive offset value enables the system to provide more customized compression processing solutions for image data with different complexity and quality requirements, ensuring both data quality and efficient transmission.

[0143] See Figure 2 As shown, in some embodiments of the present application, this embodiment provides a method for processing graphic image data based on machine learning, comprising the following steps:

[0144] S100: Determining image data to be processed, collecting image feature parameters of the image data to be processed, analyzing the image feature parameters, and determining an initial compression ratio of the image data to be processed based on the analysis results; wherein the initial compression ratio is a ratio of high-quality compression to low-quality compression of the image data to be processed;

[0145] S200: Dividing the image data to be processed into regions based on the image recognition model to obtain a plurality of monitoring regions; collecting regional feature information of each monitoring region, and determining whether to optimize the initial compression ratio based on the regional feature information;

[0146] S300: When it is determined that the initial compression ratio is to be optimized, real-time network data is collected, and an optimization coefficient of the initial compression ratio is determined based on the regional feature information and the real-time network data, and an optimized compression ratio is obtained;

[0147] S400: compressing the image data to be processed according to the optimized compression ratio, and transmitting the compressed image data to be processed.

[0148] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware. Furthermore, the present application may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0149] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems) and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 a process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0150] These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 a process or multiple processes and / or boxes Figure 1 The function specified in one or more boxes.

[0151] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing the instructions executed on the computer or other programmable device for implementing the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0152] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, ordinary technicians in the field should understand that the specific implementation methods of the present invention can still be modified or replaced by equivalents. Any modification or equivalent replacement that does not depart from the spirit and scope of the present invention should be covered by the scope of protection of the claims of the present invention.

Claims

1. A graphic image data processing system based on machine learning, characterized in that: include: an acquisition layer configured to determine image data to be processed, acquire image feature parameters of the image data to be processed, analyze the image feature parameters, and determine an initial compression ratio of the image data to be processed based on the analysis results; wherein the initial compression ratio is a ratio between high-quality compression and low-quality compression of the image data to be processed; The judgment layer is configured to divide the image data to be processed into regions based on an image recognition model to obtain a plurality of monitoring regions; collect regional feature information of each monitoring region, and determine whether to optimize the initial compression ratio based on the regional feature information; an optimization layer configured to, when determining to optimize the initial compression ratio, collect real-time network data, determine an optimization coefficient of the initial compression ratio based on the regional feature information and the real-time network data, and obtain an optimized compression ratio; The presentation layer is configured to compress the image data to be processed according to the optimized compression ratio and transmit the compressed image data to be processed; The judgment layer divides the image data to be processed into regions based on the image recognition model to obtain a number of monitoring regions, including: Using the image recognition model to divide the image data to be processed into regions to obtain a number of monitoring areas; Performing feature extraction on each of the monitoring areas to obtain regional feature information corresponding to each of the monitoring areas; When the judgment layer judges whether to optimize the initial compression ratio according to the regional feature information, it includes: Collecting regional characteristic information corresponding to each of the monitoring areas, wherein the regional characteristic information includes text density, image complexity, color richness, and key information distribution rate; Obtaining characteristic standard values ​​corresponding to the regional characteristic information, the characteristic standard values ​​including standard text density, standard image complexity, standard color richness, and standard key information distribution rate; Counting the number of regional feature information in each monitoring area that is greater than or equal to the corresponding feature standard value, and recording it as the feature number; Determine the level of the current monitoring area according to the number of features; If the number of features is less than 3, the level of the monitoring area is determined to be the first level; If the number of features is equal to 3, determining that the level of the monitoring area is the second level; If the number of features is equal to 4, determining that the level of the monitoring area is the third level; The number and value of the monitoring areas corresponding to the second level and the third level are counted, and whether to optimize the initial compression ratio is determined according to the number and value.

2. The machine learning-based image data processing system according to claim 1, characterized in that: The acquisition layer analyzes the image feature parameters and determines the initial compression ratio of the to-be-processed graphic image data based on the analysis result, including: Analyzing the image characteristic parameters to obtain the image type, transmission priority, image resolution, data volume, image noise level value, and color depth characteristic value of the image data to be processed; determining a compression type according to the image type; Constructing the compression type, transmission priority, image resolution, data volume, image noise level value and color depth feature value into an initial compression vector group; The initial compression vector group is compared with historical data, and an initial compression ratio of the to-be-processed graphic image data is determined according to the comparison result.

3. The machine learning-based image data processing system according to claim 2, characterized in that: When the acquisition layer determines the initial compression ratio of the to-be-processed graphic image data according to the comparison result, it includes: When there exists a historical compression vector group identical to the initial compression vector group in the historical data, using a historical compression ratio corresponding to the historical compression vector group as the initial compression ratio; When there is no historical compression vector group identical to the initial compression vector group in the historical data, historical compression vector groups corresponding to the same compression type and transmission priority are screened out and constructed as a comparison vector group; and matching degrees between the initial compression vector group and the comparison vector group are calculated one by one, and the initial compression ratio is determined based on the matching degrees.

4. The machine learning-based image data processing system according to claim 3, characterized in that: When the acquisition layer determines the initial compression ratio according to the matching degree, it includes: Extract the maximum value of the matching degree and record it as the maximum matching degree; Comparing the maximum matching degree with a first maximum matching degree and a second maximum matching degree, and determining the initial compression ratio according to the comparison result; wherein the first maximum matching degree is less than the second maximum matching degree; When the maximum matching degree is less than or equal to the first maximum matching degree, determining the initial compression ratio to be the first compression ratio; When the maximum matching degree is greater than the first maximum matching degree and less than or equal to the second maximum matching degree, determining the initial compression ratio to be a second compression ratio; wherein the second compression ratio is less than the first compression ratio; When the maximum matching degree is greater than the second maximum matching degree, the initial compression ratio is determined to be a third compression ratio; wherein the third compression ratio is less than the second compression ratio.

5. The machine learning-based image data processing system according to claim 4, characterized in that: When the judgment layer judges whether to optimize the initial compression ratio according to the quantity and value, it includes: Comparing the quantity and value with a quantity threshold, and determining whether to optimize the initial compression ratio based on the comparison result; When the sum of the quantity and the value is greater than the quantity threshold, determining to optimize the initial compression ratio; Otherwise, it is determined that the initial compression ratio is not to be optimized, and the initial compression ratio is used as the optimized compression ratio.

6. The machine learning-based image data processing system according to claim 5, characterized in that: The optimization layer determines the optimization coefficient of the initial compression ratio according to the regional feature information and the real-time network data, and obtains the optimized compression ratio, including: Calculating the difference between each regional characteristic information and the corresponding characteristic standard value in each of the monitoring areas corresponding to the second level and the third level, and constructing all the differences into a first set; Extracting features from the real-time network data to obtain a number of network feature values; Get the network standard value corresponding to each network characteristic value; Calculate the difference between each network characteristic value and the corresponding network standard value, and construct all the differences into a second set; Calculating a feature offset value according to the first set, calculating a network offset value according to the second set, and calculating a comprehensive offset value according to the feature offset value and the network offset value; determining an optimization coefficient of the initial compression ratio based on the comprehensive offset value; The product value of the optimization coefficient and the initial compression ratio is used as the optimized compression ratio.

7. The machine learning-based image data processing system according to claim 6, characterized in that: When the optimization layer determines the optimization coefficient of the initial compression ratio based on the comprehensive offset value, it includes: Comparing the comprehensive offset value with a first comprehensive offset value and a second comprehensive offset value, and determining an optimization coefficient of the initial compression ratio according to the comparison result; wherein the first comprehensive offset value is smaller than the second comprehensive offset value; When the comprehensive offset value is less than or equal to the first comprehensive offset value, determining the optimization coefficient of the initial compression ratio to be the first optimization coefficient; When the comprehensive offset value is greater than the first comprehensive offset value and less than or equal to the second comprehensive offset value, determining the optimization coefficient of the initial compression ratio to be a second optimization coefficient; wherein the second optimization coefficient is less than the first optimization coefficient; When the comprehensive offset value is greater than the second comprehensive offset value, the optimization coefficient of the initial compression ratio is determined to be a third optimization coefficient; wherein the third optimization coefficient is smaller than the second optimization coefficient.

8. A method for processing graphic and text image data based on machine learning, applied to the graphic and text image data processing system based on machine learning according to any one of claims 1 to 7, characterized in that: include: Determining image data to be processed, collecting image feature parameters of the image data to be processed, analyzing the image feature parameters, and determining an initial compression ratio of the image data to be processed based on the analysis results; wherein the initial compression ratio is a ratio of high-quality compression to low-quality compression of the image data to be processed; Dividing the image data to be processed into regions based on an image recognition model to obtain a plurality of monitoring regions; collecting regional feature information of each monitoring region, and determining whether to optimize the initial compression ratio based on the regional feature information; When it is determined that the initial compression ratio is to be optimized, real-time network data is collected, and an optimization coefficient of the initial compression ratio is determined based on the regional characteristic information and the real-time network data, and an optimized compression ratio is obtained; The image data to be processed is compressed according to the optimized compression ratio, and the compressed image data to be processed is transmitted.

Images