Digital education resource sharing method and system based on image contrast enhancement

Abstract

The invention discloses a digital education resource sharing method and system based on image contrast enhancement, and belongs to the technical field of digital education resource processing. Through typed feature extraction and association library construction, an accurate feature basis can be provided for subsequent adaptive enhancement, it is ensured that core details of different disciplines and types of education images are effectively identified and reserved, and different image types are adapted through feature weight vectors to realize type adaptive enhancement; a parameter range is constrained through a projection operator, noise amplification caused by excessive enhancement or detail loss caused by insufficient enhancement is avoided, and robustness can be effectively improved; initial fast enhancement and steady-state noise suppression are both considered through fractional order time-varying gain, and dynamic balance is achieved; through feature matching verification, it is ensured that the key details of the enhanced image and the original type are highly consistent; and over-enhancement or under-enhancement is avoided by using an iterative adjustment mechanism, and adaptive parameter optimization is realized.

Description

Technical Field

[0001] This invention relates to the field of digital educational resource processing technology, and specifically to a method for sharing digital educational resources based on image contrast enhancement. Background Technology

[0002] Image contrast enhancement-based digital education resource sharing is a novel resource sharing method that integrates computer vision and educational informatization technologies. Its core idea is to use image contrast enhancement technology to improve the visibility, clarity, and teaching expressiveness of image content in digital education resources, thereby improving the sharing efficiency and learning experience of resources across platforms, devices, and networks.

[0003] Image contrast enhancement techniques are typically designed for general image optimization, such as medical images and industrial inspection. However, educational resources have diverse image types, such as textbook illustrations, experimental flowcharts, and handwritten notes. Existing algorithms struggle to balance the need for detail preservation and enhancement across different disciplines. For instance, enhancing lines in mathematical formulas may disrupt the symbol structure, while color enhancement in artistic images may deviate from the original teaching intent, resulting in insufficient targeted adaptation to educational scenarios. Summary of the Invention

[0004] The purpose of this invention is to provide a digital education resource sharing method based on image contrast enhancement, which addresses the problem of insufficient targeted adaptation to educational scenarios in existing solutions.

[0005] The objective of this invention can be achieved through the following technical solutions: A method for sharing digital educational resources based on image contrast enhancement includes: Acquire diverse image sets from educational resources, automatically classify image types using pre-built classification models, extract key detail features of each image type, and construct a multimodal detail feature association library; Based on the constructed multimodal detail feature association library, a fractional-order time-varying gain adaptive enhancement model is designed. The model calls the corresponding detail feature weights according to the image type, constrains the range of enhancement parameters through projection operators, and dynamically adjusts the enhancement parameters. The input data is loaded and initialized, and the type prediction label corresponding to the input educational image to be enhanced is obtained. The type prediction label is subjected to enhancement reliability analysis. Based on the passing label in the analysis results, the designed adaptive enhancement model is used to perform comparative enhancement processing on the educational image to be enhanced. The similarity between the enhanced educational image and the corresponding type detail features is verified, and the corresponding enhancement parameters are dynamically iteratively adjusted. Finally, the enhanced educational image that meets the detail preservation requirements is used for resource sharing.

[0006] Furthermore, the diverse collection of images includes textbook illustrations, experimental flowcharts, and handwritten notes; Furthermore, a classification model was constructed using a ResNet-50 convolutional neural network, and type labels were assigned to the preprocessed images. , or ; , , These correspond to the illustration type tag, flowchart type tag, and note type tag, respectively.

[0007] Furthermore, when extracting key detail features for each type of image, Canny edge detection is used to extract line contour features for illustration-type images. ; Extracting node connection features from flowchart-type images using Hough transform. ; Extracting stroke texture features from note-type images using HOG. .

[0008] Furthermore, when constructing a multimodal detailed feature association library, the mapping between storage types and features is: key-value pairs. Store type labels and feature vectors to achieve typed storage of features; is the predicted label for the image type, and F is the key detail feature vector of the image; Calculate the cosine similarity of features of the same type and establish an association matrix to associate features; Furthermore, the database structure adopts a relational database, and the table fields include image ID, type label, feature vector, and association matrix.

[0009] Furthermore, based on the image type, the label is predicted, and the corresponding feature weight vector is retrieved from the multimodal detail feature association library. , For contrast weight, Assign a weight to the sharpening level and initialize the initial values ​​for the enhancement parameters.

[0010] Furthermore, when calculating the fractional time-varying gain, the fractional time-varying gain is designed to dynamically balance the enhancement intensity and noise suppression. The relevant expressions are as follows: ;in, For the gain growth rate, ; is the L2 norm of the current image feature vector F; μ is the gain adjustment exponent, with a value range of (0,1); θ is the gain attenuation rate, θ>0; It is a fractional differential operator, ν∈(0,1).

[0011] Furthermore, a projection operator is introduced. The constraint enhancement parameters are kept within a reasonable physical range to avoid over-enhancement or under-enhancement. The relevant expressions are as follows: ;in, These are candidate contrast parameters, which are the contrast values ​​initially calculated by the enhancement algorithm; This represents a reasonable range for contrast values. These represent the minimum and maximum contrast values, respectively. ;in, The candidate sharpening level parameter is the sharpening value initially calculated by the enhancement algorithm; To determine a reasonable range of sharpening levels, These represent the minimum and maximum sharpening values, respectively.

[0012] Furthermore, based on the obtained type prediction labels, an enhancement reliability analysis is performed on the corresponding educational images to be enhanced. The first and second enhancement impact values ​​corresponding to the type prediction labels are statistically analyzed to obtain the pass or fail labels.

[0013] Furthermore, detail feature vectors are extracted from the enhanced image; the cosine similarity formula is used to calculate the similarity between the detail feature vector and all features in the feature template library, and the maximum value is taken as the current similarity. k is the iteration number index; When performing parameter iterative adjustments, if ≥Th or k≥ If the similarity threshold is reached, the iteration stops; Th is the similarity threshold. Set the number of iterations as a threshold; otherwise, perform the following adjustments: Calculate the similarity difference ; The enhancement parameters are dynamically updated using the similarity difference.

[0014] A digital education resource sharing system based on image contrast enhancement includes: The multimodal detail feature association library processing module acquires a collection of diverse image types from educational resources, automatically classifies the image types using a pre-built classification model, extracts key detail features from each type of image, and constructs a multimodal detail feature association library. The enhancement parameter dynamic adjustment processing module, based on the constructed multimodal detail feature association library, designs a fractional-order time-varying gain adaptive enhancement model, calls the corresponding detail feature weights according to the image type, constrains the range of enhancement parameters through projection operators, and dynamically adjusts the enhancement parameters. The verification analysis and enhancement processing module loads and initializes the input data, obtains the type prediction label corresponding to the input educational image to be enhanced, performs enhancement reliability analysis on the type prediction label, and uses the pass label in the analysis results to perform comparative enhancement processing on the educational image to be enhanced using the designed adaptive enhancement model. It verifies the similarity between the enhanced educational image and the corresponding type detail features, and dynamically iterates and adjusts the corresponding enhancement parameters. Finally, the enhanced educational image that meets the detail preservation requirements is used for resource sharing.

[0015] Compared to existing solutions, the beneficial effects achieved by this invention are: This invention provides a precise feature foundation for subsequent adaptive enhancement through typified feature extraction and association library construction, ensuring that the core details of educational images of different disciplines and types are effectively identified and preserved. It solves the problem that general enhancement algorithms cannot be optimized for the diversity of educational images, and improves the targeting and accuracy of detail preservation.

[0016] This invention achieves type-adaptive enhancement by adapting feature weight vectors to different image types; it avoids noise amplification due to over-enhancement or detail loss due to under-enhancement by constraining the parameter range through projection operators, thus effectively improving robustness; it achieves dynamic balance by balancing initial rapid enhancement and steady-state noise suppression through fractional time-varying gain; and it ensures that key details are enhanced in a targeted manner by associating feature weights with enhancement parameters.

[0017] This invention ensures that the enhanced image is highly consistent with the key details of the original type through feature matching verification; it avoids over-enhancement or under-enhancement by using an iterative adjustment mechanism to achieve adaptive parameter optimization; and it controls the process by setting a preset threshold and a maximum number of iterations, balancing processing efficiency and quality, which can effectively improve robustness. Enhanced images that meet the requirements can be directly used for educational resource sharing, improving user experience and learning efficiency. Attached Figure Description

[0018] The invention will now be further described with reference to the accompanying drawings.

[0019] Figure 1 This is a flowchart illustrating the steps of implementing a digital education resource sharing method based on image contrast enhancement according to the present invention. Figure 2 This is a block diagram of a digital education resource sharing system based on image contrast enhancement according to the present invention. Detailed Implementation

[0020] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0021] Example 1, as Figure 1 As shown, this invention is a method for sharing digital educational resources based on image contrast enhancement, comprising: The process involves acquiring diverse image sets from educational resources, automatically classifying image types using a pre-built classification model, extracting key detail features from each image type, and constructing a multimodal detail feature association library. Specific steps include: The diverse image collection includes textbook illustrations, experimental flowcharts, and handwritten notes; it can be supplemented according to the application needs of actual application scenarios. Each image undergoes preprocessing, including but not limited to grayscale conversion, Gaussian filtering for noise reduction, and size normalization. These are all existing conventional technical solutions, and the specific implementation steps will not be elaborated here. A classification model is constructed using a ResNet-50 convolutional neural network, which is a conventional existing technique. The specific implementation steps are not detailed here. Type labels are then assigned to the preprocessed images. , or ; , , These correspond to the illustration type tag, flowchart type tag, and note type tag, respectively. The preprocessed images are divided into training, validation, and test sets in a 7:2:1 ratio, and trained using the cross-entropy loss function. The relevant expression is as follows: Where i is the index of different label types, i=1,2,3; One-hot encoding for the actual label; To predict class probabilities; When extracting key detail features from various image types, such as the line outlines of illustrations, the node connections in flowcharts, and the stroke textures of handwritten notes, Canny edge detection is used to extract the line outline features of illustration-type images. The expression involved is: ;in, This is an edge detection function; A grayscale image is a color image converted into a single-channel grayscale form, where each pixel value represents the brightness information of the image; threshold1 is the low threshold for edge detection; threshold2 is the high threshold for edge detection. Extracting node connection features from flowchart-type images using Hough transform. The expression involved is: ;in, Here, ρ is the probabilistic Hough transform function; ρ is the distance resolution; θ is the angular resolution; threshold is the accumulator threshold; minLineLength is the minimum line length; maxLineGap is the maximum line gap. Extracting stroke texture features from note-type images using HOG. The expression involved is: Where orientations is the number of directional bins; Cell unit size; This represents the number of cell units within the block. When building a multimodal detailed feature association library, the mapping between storage type and features is as follows: key-value pairs Store type labels and feature vectors to achieve typed storage of features; is the predicted label for the image type, and F is the key detail feature vector of the image; Calculate the cosine similarity of similar features, establish an association matrix for feature association, and the relevant expression is: ;in, These are the element values ​​in the correlation matrix; These are two feature vectors from images of the same type; For dot product, It is an L2 norm; Furthermore, the database structure uses a relational database, with table fields including image ID, type label, feature vector, and association matrix. For example, in the feature library for textbook illustrations, Two feature vectors with high values ​​correspond to visually similar illustration contours, which can be used to quickly match or cluster similar teaching resources.

[0022] In this embodiment of the invention, by extracting typological features and constructing an association library, a precise feature foundation can be provided for subsequent adaptive enhancement, ensuring that the core details of educational images of different disciplines and types are effectively identified and preserved. Core details include illustration lines, flowchart nodes, and handwritten strokes. This solves the problem that general enhancement algorithms cannot be optimized for the diversity of educational images, and improves the targeting and accuracy of detail preservation.

[0023] Based on the constructed multimodal detail feature association library, a fractional-order time-varying gain adaptive enhancement model is designed. It calls the corresponding detail feature weights according to the image type, constrains the range of enhancement parameters through projection operators, and dynamically adjusts the enhancement parameters. The specific steps include: Predict the label based on the image type, and retrieve the corresponding feature weight vector from the multimodal detail feature association library. , For contrast weight, Weight the sharpening level and initialize the enhancement parameters, including the initial contrast value. Initial value of sharpening ; When calculating the fractional time-varying gain, the fractional time-varying gain γ(t) is designed, and the enhancement strength and noise suppression are dynamically balanced. The relevant expressions are as follows: ;in, For the gain growth rate, The default value is 0.7, which controls the rate at which the gain increases with the magnitude of the eigenvector. is the L2 norm of the current image feature vector F, reflecting the richness of image details; μ is the gain adjustment exponent, with a value range of (0,1), and a default value of 0.6, balancing detail sensitivity and robustness; θ is the gain decay rate, θ>0, with a default value of 0.2, to prevent unbounded gain growth; For fractional differential operators, ν∈(0,1), with a default value of 0.3. ;in, Let f(t) represent the νth order fractional derivative of the function f(t); ν is the Gamma function; f(t) is the fractional derivative order; f(t) is the differentiable function. Let t be the first integer derivative of the function f at a historical time σ; σ is the integration variable; t is the current time. The value is ν raised to the power of the time difference between the current time t and the historical time σ. It should be noted that the gain growth rate This is used to control the rate at which the gain increases with the magnitude of the eigenvector, reflecting the initial enhancement of image detail richness. When the value is increased, the enhancement intensity is significantly improved, and useful details, such as the outlines of textbook illustrations and formulas, will stand out more. However, noise details, such as grainy noise, will also be amplified. Therefore, for scenarios with high detail requirements, such as experimental flowcharts, the value range is specified to 0.8~1.0; for scenarios with low noise requirements, such as low-resolution handwritten notes, the value range is specified to 0.4~0.6; for balanced scenarios, the default value is 0.7, and the specific value can be customized according to the needs of the actual application scenario. The gain adjustment index μ is used to balance detail sensitivity and robustness, controlling the nonlinear response of the gain to the magnitude of the feature vector. When μ increases, detail sensitivity improves, and the discrimination of useful details is higher, but noise suppression weakens. When μ decreases, robustness is enhanced, but the enhancement of useful details is weakened. Therefore, for scenarios with strong detail discrimination, such as mathematical formula recognition, the value range is 0.7~0.9; for high-noise scenarios, such as scanned documents, the value range is 0.3~0.5; for general scenarios, the default value is 0.6. The gain attenuation rate θ is used to prevent unbounded gain growth and suppresses excessive enhancement through fractional derivative terms. As θ increases, noise suppression capability improves and enhancement intensity decreases. If attenuation is excessive, the contrast of useful details will decrease. Therefore, for high-noise scenarios, the value range is specifically 0.3~0.5; for low-noise scenarios, the value range is specifically 0.1~0.2; to prevent over-enhancement, the value range is specifically ≥0.2 to avoid gain runaway. The fractional derivative order ν reflects the weight of the influence of historical feature changes on the current gain, controlling the dynamic adaptability of the enhancement; increasing ν improves the dynamic response but slightly weakens noise suppression; decreasing ν enhances smoothness but slows the dynamic response; therefore, for dynamic detail scenarios, such as experimental animation frames, the value range is specifically 0.6~0.8; for static detail scenarios, such as textbook illustrations, the value range is specifically 0.4~0.5; for low-noise smoothing scenarios, the default value is 0.3. In addition, the projection operator is introduced. The constraint enhancement parameters are kept within a reasonable physical range to avoid over-enhancement or under-enhancement. The relevant expressions are as follows: ;in, These are candidate contrast parameters, which are the contrast values ​​initially calculated by the enhancement algorithm; The appropriate range of contrast values ​​is preset by the user or the system based on the image type and application scenario. These are the minimum and maximum contrast values, respectively, with default values ​​of 0.5 and 2; C is the final contrast parameter, which is the effective contrast value after projection constraints. ;in, The candidate sharpening level parameter is the sharpening value initially calculated by the enhancement algorithm; The appropriate range of sharpening values ​​is preset by the user or the system based on the image type and application scenario. These are the minimum and maximum sharpening values, respectively, with default values ​​of 0 and 1; S is the final sharpening parameter, which is the effective sharpening value after projection constraints. Furthermore, the calculation rules for the projection operator proj(x,[a,b]) are as follows: if the candidate parameter is within the interval, it is directly retained; if the candidate parameter is less than the lower limit of the interval, the lower limit value is taken; if it is greater than the upper limit of the interval, the upper limit value is taken; the relevant expressions are: ; It should be noted that the projection operator effectively avoids the problem of enhancement parameters going out of bounds by mapping candidate parameters to a preset range, ensuring that the image enhancement effect meets visual requirements without compromising image quality. The coordinated work of the parameters achieves controllability and robustness in the enhancement process.

[0024] When dynamically adjusting enhancement parameters and outputting results, the constrained parameters are applied to the original image to generate the enhanced image: ; where Adjust() is the enhancement function, which adjusts the pixel brightness distribution according to the contrast C and enhances edge details according to the sharpening degree S.

[0025] In this embodiment of the invention, feature weight vectors are adapted to different image types. For example, flowcharts prioritize enhancing line connections, while handwritten notes prioritize enhancing stroke texture, achieving type-adaptive enhancement. By constraining the parameter range through projection operators, noise amplification due to over-enhancement or detail loss due to under-enhancement is avoided, effectively improving robustness. A dynamic balance is achieved by using fractional time-varying gain to balance initial rapid enhancement and steady-state noise suppression. The correlation between feature weights and enhancement parameters ensures that key details are enhanced in a targeted manner, such as lines in textbook illustrations or strokes in handwritten notes.

[0026] The input data is loaded and initialized, and the type prediction labels corresponding to the input educational image to be enhanced are obtained. Enhancement reliability analysis is performed on the type prediction labels. Based on the passing labels in the analysis results, the designed adaptive enhancement model is used to perform comparative enhancement processing on the educational image to be enhanced. The similarity between the enhanced educational image and the corresponding type detail features is verified, and the corresponding enhancement parameters are dynamically and iteratively adjusted. Finally, the enhanced educational images that meet the detail preservation requirements are used for resource sharing. The specific steps include: When loading and initializing the input data, the data to be processed is loaded, including the input educational image to be enhanced. Obtain its type prediction label For example, illustration type tags Flowchart type tags Or note type tags ; Based on the obtained type prediction labels, perform enhancement reliability analysis on the educational images to be enhanced, and statistically analyze the first and second enhancement impact values ​​corresponding to the type prediction labels. If the first reinforcement impact value is less than or equal to the first reinforcement impact threshold, and the second reinforcement impact value is less than or equal to the second reinforcement impact threshold, then a pass label is generated. The first reinforcement impact threshold and the second reinforcement impact threshold can be determined based on the simulation test results of the previous sample data, such as the median of all first reinforcement impact values ​​and the median of all second reinforcement impact values ​​obtained from the simulation test. They can also be customized and adjusted by professionals in this field according to the application requirements and specifications of the actual application scenario. The specific values ​​are not limited. Conversely, if a failure label is generated, subsequent processing of the associated educational image to be enhanced will be stopped based on the failure label, and maintenance personnel will be alerted to intervene. The steps for obtaining the first enhancement influence value and the second enhancement influence value include: The statistical type prediction label corresponds to the historical total number of processed cases, and the statistical feedback label indicates the total number of first and second anomalies. The values ​​of the historical total number of processed cases, the total number of first anomalies, and the total number of second anomalies corresponding to the type prediction label are extracted. The corresponding first and second reinforcement impact values ​​are then calculated using the following expressions: ; Wherein, QY1 and QY2 are the first and second enhancement impact values, respectively; NC, ZY1, and ZY2 are the total number of historically processed, the total number of first anomalies, and the total number of second anomalies corresponding to the type prediction label, respectively. It should be noted that the first anomaly is that the display effect of the predicted label of the type after the augmented education image processing, according to the statistical manual feedback, has not returned to normal. That is, there were anomalies such as incomplete or unclear key details before the augmented education image processing. The second anomaly is specifically the display effect of the type prediction label after the augmented educational image processing, which is abnormal compared to the display effect of the type prediction label before the augmented educational image processing, based on manual feedback. For example, the key detail features are normal before processing, but incomplete or unclear key detail features after processing. The negative impact of the second anomaly is greater than that of the first anomaly. The first enhanced impact value serves to supervise and digitize the handling of abnormal display effects, and to determine the degree of enhanced impact of the abnormal display effect handling dimension corresponding to the prediction label of the type. The second enhanced impact value serves to monitor and digitize the normal display effect processing dimension, and determine the degree of enhanced impact of the predicted label of the type corresponding to the normal display effect processing dimension. Furthermore, the calculation of the first enhancement influence value and the second enhancement influence value in the embodiments of the present invention can also be performed by training existing neural network algorithms, such as convolutional neural networks, recurrent neural networks, and graph neural networks, using the total number of samples processed, the total number of first anomalies, and the total number of second anomalies. The trained model is then used to perform data analysis on the total number of samples processed in history corresponding to the type prediction label and the total number of first and second anomalies marked by statistical feedback, and the output values ​​are marked as the first enhancement influence value and the second enhancement influence value, respectively. These are all existing conventional technical solutions, and the specific implementation steps are not described here. It should be explained that by performing enhancement reliability analysis on the educational images to be enhanced and dynamically implementing contrast enhancement processing based on the analysis results, we can avoid performing contrast enhancement processing on all types of predicted labels that do not meet the requirements for contrast enhancement processing, which would waste processing data resources and affect the overall enhancement effect. This can effectively improve the reliability and resource utilization of contrast enhancement processing for different types of predicted labels.

[0027] Call the corresponding type of feature template set from the multimodal detailed feature association library. For example, the Canny line outline feature library for textbook illustrations; Additionally, the similarity threshold Th=0.85, the maximum number of iterations Nmax=5, and the initial value of the iteration counter k=0 are set; where the similarity threshold is used to determine whether the detail preservation meets the standard; the maximum number of iterations avoids infinite loops; When performing contrast enhancement processing, the educational image to be enhanced is... and type prediction labels Input the designed fractional-order time-varying gain adaptive enhancement model to generate the enhanced image for the current iteration. ; The model automatically calls the corresponding feature weights, such as the contour weights of textbook illustrations, and constrains the enhancement parameters within a preset range through projection operators; for example, contrast C∈[0.5,2.0] and sharpness S∈[0.0,1.0]. Enhanced image Extracting detail feature vectors The cosine similarity formula is used to calculate the detailed feature vector. The similarity score is calculated by taking the maximum similarity score with all features in the feature template library. The expression involved is as follows: ;in, This represents the maximum similarity value after k iterations; k is the iteration index. The i-th feature vector in the feature template library; It is the dot product of two vectors, reflecting the consistency of feature directions; These are the L2 norms of the enhanced feature vector and the template feature vector, respectively, used for normalization; A set of feature templates for the corresponding image type; It should be noted that this formula calculates feature similarity through directional consistency and normalization, and combines maximum value screening to ensure that detail preservation meets type requirements, providing a quantitative standard for quality verification of educational image enhancement; each parameter serves the core goals of accurate matching and detail preservation, ensuring that the enhanced image meets visual requirements without losing key information.

[0028] When performing parameter iterative adjustments, if ≥Th or k≥ If the similarity threshold is reached, the iteration stops; Th is the similarity threshold. Set the number of iterations as a threshold; otherwise, perform the following adjustments: Calculate the similarity difference ; When dynamically updating enhancement parameters using similarity differences, the relevant expression is: ; ;in, The updated contrast parameters; The contrast parameter from the previous iteration; The updated sharpening level parameters; This represents the sharpening level parameter from the previous iteration; α is the contrast adjustment coefficient, with a default value of 0.2; β is the sharpening adjustment coefficient, with a default value of 0.15. It should be noted that this formula drives parameter adjustment through similarity difference and combines it with the constraint range of the projection operator to achieve dynamic optimization of enhancement parameters, ensuring that the image retains key details while maintaining good visual quality during the iteration process. The design of each parameter serves the goal of balancing detail preservation and image naturalness; Furthermore, unlike existing technical solutions that only implement verification and supervision through a single regulatory scheme, this embodiment of the invention enhances the reliability analysis of type prediction labels and coordinates it with the similarity between the enhanced educational image and the corresponding type detail features, which can effectively improve the diversity and reliability of the enhanced processing effect supervision.

[0029] Enhanced images that meet the conditions The images are stored in an educational resource library, and information such as image type, enhancement parameters, and similarity value are recorded for resource retrieval and sharing, such as the textbook illustration library of an online education platform.

[0030] In this embodiment of the invention, feature matching verification ensures that the enhanced image is highly consistent with the key details of the original type; an iterative adjustment mechanism is used to avoid over-enhancement, such as excessive contrast leading to overexposure, or under-enhancement, such as insufficient sharpening leading to blurring, thereby achieving adaptive parameter optimization; preset thresholds and maximum number of iterations control the process, balancing processing efficiency and quality, which can effectively improve robustness. Enhanced images that meet the requirements can be directly used for sharing educational resources, improving user experience and learning efficiency; Through the synergy of the above steps, a closed-loop control of educational images from enhancement to quality verification is achieved, which can provide high-quality, detailed visual materials for resource sharing.

[0031] Example 2, as Figure 2 As shown, a digital education resource sharing system based on image contrast enhancement includes: The multimodal detail feature association library processing module acquires a collection of diverse image types from educational resources, automatically classifies the image types using a pre-built classification model, extracts key detail features from each type of image, and constructs a multimodal detail feature association library. The enhancement parameter dynamic adjustment processing module, based on the constructed multimodal detail feature association library, designs a fractional-order time-varying gain adaptive enhancement model, calls the corresponding detail feature weights according to the image type, constrains the range of enhancement parameters through projection operators, and dynamically adjusts the enhancement parameters. The verification analysis and enhancement processing module loads and initializes the input data, obtains the type prediction label corresponding to the input educational image to be enhanced, performs enhancement reliability analysis on the type prediction label, and uses the pass label in the analysis results to perform comparative enhancement processing on the educational image to be enhanced using the designed adaptive enhancement model. It verifies the similarity between the enhanced educational image and the corresponding type detail features, and dynamically iterates and adjusts the corresponding enhancement parameters. Finally, the enhanced educational image that meets the detail preservation requirements is used for resource sharing.

[0032] In the several embodiments provided by this invention, it should be understood that the disclosed system can be implemented in other ways. For example, the embodiments of the invention described above are merely illustrative; for example, the division of modules is only a logical functional division, and there may be other division methods in actual implementation.

[0033] The modules described as separate components may or may not be physically separate. The components shown as modules may or may not be physical modules; they may be located in one place or distributed across multiple network modules. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.

[0034] Furthermore, the functional modules in the various embodiments of the present invention can be integrated into one processing module, or each module can exist physically separately, or two or more modules can be integrated into one module. The integrated module can be implemented in hardware or in the form of hardware plus software functional modules.

[0035] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the essential characteristics of the present invention.

[0036] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A method for sharing digital educational resources based on image contrast enhancement, characterized in that, include: Acquire diverse image sets from educational resources, automatically classify image types using pre-built classification models, extract key detail features of each image type, and construct a multimodal detail feature association library; Based on the constructed multimodal detail feature association library, a fractional-order time-varying gain adaptive enhancement model is designed. The model calls the corresponding detail feature weights according to the image type, constrains the range of enhancement parameters through projection operators, and dynamically adjusts the enhancement parameters. The input data is loaded and initialized, and the type prediction label corresponding to the input educational image to be enhanced is obtained. The type prediction label is subjected to enhancement reliability analysis. Based on the passing label in the analysis results, the designed adaptive enhancement model is used to perform comparative enhancement processing on the educational image to be enhanced. The similarity between the enhanced educational image and the corresponding type detail features is verified, and the corresponding enhancement parameters are dynamically iteratively adjusted. Finally, the enhanced educational image that meets the detail preservation requirements is used for resource sharing.

2. The digital education resource sharing method based on image contrast enhancement according to claim 1, characterized in that, The diverse collection of images includes textbook illustrations, experimental flowcharts, and handwritten notes; Furthermore, a classification model was constructed using a ResNet-50 convolutional neural network, and type labels were assigned to the preprocessed images. , or ; , , These correspond to the illustration type tag, flowchart type tag, and note type tag, respectively.

3. The digital education resource sharing method based on image contrast enhancement according to claim 2, characterized in that, When extracting key detail features from various image types, Canny edge detection is used to extract line contour features from illustration-type images. ; Extracting node connection features from flowchart-type images using Hough transform. ; Extracting stroke texture features from note-type images using HOG. .

4. The digital educational resource sharing method based on image contrast enhancement according to claim 3, characterized in that, When building a multimodal detailed feature association library, the mapping between storage type and features is: key-value pairs. Store type labels and feature vectors to achieve typed storage of features; is the predicted label for the image type, and F is the key detail feature vector of the image; Calculate the cosine similarity of features of the same type and establish an association matrix to associate features; Furthermore, the database structure adopts a relational database, and the table fields include image ID, type label, feature vector, and association matrix.

5. A digital education resource sharing method based on image contrast enhancement according to claim 4, characterized in that, Predict the label based on the image type, and retrieve the corresponding feature weight vector from the multimodal detail feature association library. , For contrast weights, Assign a weight to the sharpening level and initialize the initial values ​​for the enhancement parameters.

6. The digital educational resource sharing method based on image contrast enhancement according to claim 5, characterized in that, When calculating fractional-order time-varying gain, the fractional-order time-varying gain is designed, and the enhancement strength and noise suppression are dynamically balanced. The relevant expressions are: ;in, For the gain growth rate, ; is the L2 norm of the current image feature vector F; μ is the gain adjustment exponent, with a value range of (0,1); θ is the gain attenuation rate, θ>0; It is a fractional differential operator, ν∈(0,1).

7. A method for sharing digital educational resources based on image contrast enhancement according to claim 6, characterized in that, Introducing the projection operator The constraint enhancement parameters are kept within a reasonable physical range to avoid over-enhancement or under-enhancement. The relevant expressions are as follows: ;in, These are candidate contrast parameters, which are the contrast values ​​initially calculated by the enhancement algorithm; This represents a reasonable range for contrast values. These represent the minimum and maximum contrast values, respectively. ;in, The candidate sharpening level parameter is the sharpening value initially calculated by the enhancement algorithm; To determine a reasonable range of sharpening levels, These represent the minimum and maximum sharpening values, respectively.

8. A method for sharing digital educational resources based on image contrast enhancement according to claim 7, characterized in that, Based on the obtained type prediction labels, an enhancement reliability analysis is performed on the corresponding educational images to be enhanced. The first and second enhancement impact values ​​corresponding to the type prediction labels are statistically analyzed to obtain the pass or fail labels.

9. A method for sharing digital educational resources based on image contrast enhancement according to claim 8, characterized in that, Detail feature vectors are extracted from the enhanced image; the cosine similarity formula is used to calculate the similarity between the detail feature vector and all features in the feature template library, and the maximum value is taken as the current similarity. k is the iteration number index; When performing parameter iterative adjustments, if ≥Th or k≥ If the similarity threshold is reached, the iteration stops; Th is the similarity threshold. Set the number of iterations as a threshold; otherwise, perform the following adjustments: Calculate the similarity difference ; The enhancement parameters are dynamically updated using the similarity difference.

10. A digital educational resource sharing system based on image contrast enhancement, characterized in that, include: The multimodal detail feature association library processing module acquires a collection of diverse image types from educational resources, automatically classifies the image types using a pre-built classification model, extracts key detail features from each type of image, and constructs a multimodal detail feature association library. The enhancement parameter dynamic adjustment processing module, based on the constructed multimodal detail feature association library, designs a fractional-order time-varying gain adaptive enhancement model, calls the corresponding detail feature weights according to the image type, constrains the range of enhancement parameters through projection operators, and dynamically adjusts the enhancement parameters. The verification analysis and enhancement processing module loads and initializes the input data, obtains the type prediction label corresponding to the input educational image to be enhanced, performs enhancement reliability analysis on the type prediction label, and uses the pass label in the analysis results to perform comparative enhancement processing on the educational image to be enhanced using the designed adaptive enhancement model. It verifies the similarity between the enhanced educational image and the corresponding type detail features, and dynamically iterates and adjusts the corresponding enhancement parameters. Finally, the enhanced educational image that meets the detail preservation requirements is used for resource sharing.

Images