A computer image analysis method based on big data

Abstract

The present invention relates to the technical field of image analysis, and discloses a computer image analysis method based on big data, comprising a first acquisition module, a second acquisition module, a data processing module, a first calculation module, a second calculation module, a first analysis module, a second analysis module and an execution module. Compared with traditional technologies, the present method realizes multi-source data driving, dynamic threshold determination and hierarchical enhancement strategy through the collaborative work of multiple modules, avoids the problems of over-enhancement and under-enhancement, and significantly improves the reliability and intelligence level of image enhancement. The present method has strong adaptability in different scenarios, optimizes the accuracy and efficiency of image quality processing, and provides a scientific and efficient solution for the application of computer vision technology.

Description

Technical Field

[0001] The present invention relates to the technical field of image analysis, and in particular to a computer image analysis method based on big data. Background Art

[0002] The task of image enhancement is to adjust the clarity, contrast and noise level of an image to make it more consistent with human visual perception needs or the input requirements of a specific task. In recent years, with the development of data processing capabilities and image sensing technology, intelligent analysis methods based on multi-source data have gradually become an important breakthrough in the field of image enhancement, providing more adaptive and efficient enhancement solutions through accurate collection of image information and multi-dimensional characteristic analysis.

[0003] Current image enhancement methods mostly use fixed rules or a single algorithm. Although they can improve image quality under certain specific conditions, they have many limitations when facing complex scenes. Traditional methods lack a comprehensive analysis of the details of image characteristics and cannot effectively deal with complex and diverse noise types and dynamically changing lighting environments. In the enhancement process, there is a lack of refined control over the enhancement magnitude, which easily leads to over-enhancement or under-enhancement. In addition, traditional methods pay insufficient attention to the layered characteristics of image data and fail to fully utilize the correlation between main information and sub-information, thereby limiting the comprehensiveness and adaptability of the enhancement effect, which directly affects the efficiency and quality of image processing in practical applications, especially when optimizing image clarity in complex scenes.

[0004] So we proposed a computer image analysis method based on big data to solve the above problems. Summary of the invention

[0005] The purpose of the present invention is to provide a computer image analysis method based on big data to solve the problems that the traditional methods proposed in the above background technology lack comprehensive analysis of image characteristic details, cannot effectively cope with complex and diverse noise types and dynamically changing lighting environments; lack refined control over the enhancement level during the enhancement process, and easily lead to over-enhancement or under-enhancement.

[0006] To achieve the above object, the present invention provides the following technical solution: a computer image analysis method based on big data, comprising a first acquisition module, a second acquisition module, a data processing module, a first calculation module, a second calculation module, a first analysis module, a second analysis module and an execution module, and the specific steps are as follows:

[0007] S1. The main information of the current image is collected through the first collection module, and the relevant corresponding data is retrieved from the cloud platform to form a first multi-source data set; at the same time, the second collection module collects the sub-information of the image and integrates it into a second multi-source data set;

[0008] S2. Preprocessing and dimensionlessizing the first multi-source data set and the second multi-source data set through a data processing module to ensure data standardization and facilitate subsequent calculations. The preprocessed data is further classified to generate a first data set, a second data set, a third data set, a fourth data set, a fifth data set, and a sixth data set, covering information such as image quality, resolution, and texture complexity;

[0009] S3, integrating and calculating the first data set, the second data set and the third data set through the first calculation module to generate a clarity recognition coefficient QXD;

[0010] S4. By using the first analysis module, the generated clarity recognition coefficient QXD is compared with a preset first threshold value Y to generate a first comparison result. If QXD is less than Y, it indicates that the current image needs to be enhanced, and the next step is entered; if QXD is greater than or equal to Y, it indicates that the current image does not need to be enhanced, and the method ends.

[0011] S5. If enhancement is required, the second calculation module will generate an enhancement magnitude coefficient ZQL by integrating the fourth data set, the fifth data set and the sixth data set;

[0012] S6. Through the second analysis module, the enhancement magnitude coefficient ZQL is compared with the preset second threshold R to generate a second comparison result. According to the result, corresponding image enhancement operations are performed, such as low-intensity Laplace filtering, detail layer extraction, contrast adaptive equalization, etc. Finally, the corresponding enhancement strategy is applied through the execution module to complete the image optimization.

[0013] Preferably, the first acquisition module includes an image quality main information acquisition unit, an image technical information main acquisition unit and an image content characteristic main acquisition unit;

[0014] The image quality main information acquisition unit is used to acquire noise intensity, contrast range and edge clarity;

[0015] The image technical information main acquisition unit is used to acquire image resolution, sampling rate and zoom ratio;

[0016] The image content characteristic main acquisition unit is used to acquire texture complexity, illumination uniformity and dynamic range;

[0017] The noise intensity, contrast range, edge definition, image resolution, sampling rate, scaling ratio, texture complexity, illumination uniformity, and dynamic range constitute a first multi-source data set.

[0018] Preferably, the second acquisition module includes an image quality sub-information acquisition unit, an image technical information sub-acquisition unit and an image content characteristic sub-acquisition unit;

[0019] The image quality sub-information acquisition unit is used to acquire noise variance, Gaussian noise ratio and signal-to-noise ratio;

[0020] The image technical information sub-collection unit is used to collect pixel spacing, sampling depth and interpolation error;

[0021] The image content characteristic sub-collection unit is used to collect the intensity of high-frequency components, the ratio of edge pixels and the local illumination deviation;

[0022] The noise variance, Gaussian noise proportion, signal-to-noise ratio, pixel spacing, sampling depth, interpolation error, high-frequency component intensity, edge pixel ratio and local illumination deviation constitute a second multi-source data set.

[0023] Preferably, the data processing module is used to preprocess and dimensionlessly transform the first multi-source data set and the second multi-source data set, and divide the processed first multi-source data set and the second multi-source data set into a first data set, a second data set, a third data set, a fourth data set, a fifth data set and a sixth data set;

[0024] The first data set includes noise intensity, contrast range, and edge definition;

[0025] The noise intensity is divided into actual image noise intensity A1 and required noise intensity A2;

[0026] The contrast range is divided into the actual contrast range B1 and the required contrast range B2;

[0027] Edge clarity is divided into actual edge clarity C1 and required edge clarity C2;

[0028] The second data set includes image resolution, sampling rate, and scaling ratio;

[0029] Image resolution is divided into actual image resolution D1 and required image resolution D2;

[0030] The sampling rate is divided into actual sampling rate E1 and required sampling rate E2;

[0031] The scaling ratio is divided into the actual scaling ratio F1 and the required scaling ratio F2;

[0032] The third dataset includes texture complexity, lighting uniformity, and dynamic range;

[0033] Texture complexity is divided into actual texture complexity G1 and required texture complexity G2;

[0034] Light uniformity is divided into actual light uniformity H1 and required light uniformity H2;

[0035] The dynamic range structure is divided into the actual dynamic range structure I1 and the required dynamic range structure I2;

[0036] The fourth data set includes noise variance J, Gaussian noise weight K, and signal-to-noise ratio L;

[0037] The fifth data set includes a pixel spacing M, a sampling depth N, and an interpolation error O;

[0038] The sixth data set includes high-frequency component intensity P, edge pixel ratio Q, and local illumination deviation S. Preferably, the first calculation module calculates and obtains the clarity recognition coefficient QXD by the following formula:

[0039] QXD=a1×Z1+a2×Z2+a3×Z3;

[0040] Wherein: Z1 is the first main reference coefficient, which is calculated by integrating the first data set, Z2 is the second main reference coefficient, which is calculated by integrating the second data set, Z3 is the third main reference coefficient, which is calculated by integrating the third data set, a1, a2 and a3 are weight values, and the values ​​of a1, a2 and a3 are adjusted and set by the user.

[0041] Preferably, the first calculation module calculates and obtains the first main reference coefficient Z1, the second main reference coefficient Z2 and the third main reference coefficient Z3 respectively by the following formulas:

[0042]

[0043]

[0044] Wherein: A1, B1, C1, D1, E1, F1, G1, H1 and I1 are the actual noise intensity, actual contrast range, actual edge definition, actual image resolution, actual sampling rate, actual scaling ratio, actual texture complexity, actual illumination uniformity and actual dynamic range respectively;

[0045] A2, B2, C2, D2, E2, F2, G2, H2 and I2 are the required noise intensity, required contrast range, required edge clarity, required image resolution, required sampling rate, required scaling ratio, required texture complexity, required illumination uniformity and required dynamic range, respectively;

[0046] c1, c2, c3 are weight values, and the values ​​of c1, c2 and c3 are adjusted and set by the user.

[0047] Preferably, the second calculation module calculates and obtains the enhancement magnitude coefficient ZQL by the following formula:

[0048] ZQL=b1×F1+b2×F2+b3×F3;

[0049] Where: F1 is the first sub-reference coefficient, obtained by integrating the fourth data set; F2 is the second sub-reference coefficient, obtained by integrating the fifth data set; F3 is the third sub-reference coefficient, obtained by integrating the sixth data set; b1, b2, and b3 are weight values, and the values of b1, b2, and b3 are adjusted and set by the user.

[0050] Preferably, the second calculation module calculates and obtains the first sub-reference coefficient F1, the second sub-reference coefficient F2, and the third sub-reference coefficient F3 through the following formulas respectively:

[0051]

[0052] Where: L is the signal-to-noise ratio, J is the noise variance, K is the Gaussian noise proportion, M is the pixel pitch, N is the sampling depth, O is the interpolation error, P is the high-frequency component intensity, Q is the edge pixel ratio, and S is the local illumination deviation.

[0053] Preferably, the first comparison result generated by the first analysis module is as follows:

[0054] When QXD < Y, it represents that the current image needs to be enhanced; when QXD ≥ Y, it represents that the current image does not need to be enhanced.

[0055] Preferably, the second comparison result generated by the second analysis module is specifically as follows:

[0056] When R × 95% < ZQL ≤ R, it represents that the current picture needs to be enhanced at the first level. Apply a low-intensity Laplacian filter to the current picture and enhance the edge sharpness. According to the bright and dark areas of the picture, locally increase the brightness by about 5% - 10%, and use a median filter to slightly remove noise to prevent noise diffusion caused by enhanced edges;

[0057] When R × 85% < ZQL ≤ R × 95%, it represents that the current picture needs to be enhanced at the second level. Use a multi-scale Gaussian pyramid to extract the detail layer of the image for the current picture, enhance its texture features, based on the contrast-limited adaptive histogram equalization algorithm, enhance the local contrast, and uniformize the brightness distribution. Combine with a bilateral filter to reduce the interference of noise on edge details;

[0058] When ZQL ≤ R × 85%, it represents that the current picture needs to be enhanced at the third level. Use a blind deconvolution algorithm to estimate the blur kernel and restore the current picture, globally apply Gamma correction to enhance the brightness and contrast, and use non-local means denoising to significantly reduce noise interference.

[0059] Compared with the prior art, the beneficial effects of the present invention are:

[0060] 1. Compared with traditional technologies, this method achieves multi-source data drive, dynamic threshold determination and hierarchical enhancement strategy through the collaborative work of multiple modules, avoiding the problems of over-enhancement and under-enhancement, and significantly improving the reliability and intelligence level of image enhancement. This method has strong adaptability in different scenarios, optimizes the accuracy and efficiency of image quality processing, and provides a scientific and efficient solution for the application of computer vision technology.

[0061] 2. The first acquisition module comprehensively acquires the main information of the image from multiple dimensions by introducing the image quality main information acquisition unit, the image technical information main acquisition unit and the image content characteristic main acquisition unit. This multi-dimensional acquisition breaks through the limitations of traditional single indicator acquisition and more comprehensively reflects the overall characteristics of the image. By integrating noise intensity, contrast range, edge clarity, image resolution, sampling rate, scaling ratio, texture complexity, illumination uniformity and dynamic range into the first multi-source data set, the data is not only structured, but also efficiently analyzed and utilized. The module design incorporates a variety of key image characteristics into the acquisition scope, has a wide range of applicability, can adapt to different types of image scenes and requirements, and ensures the flexibility and adaptability of the method processing. By setting up the image quality sub-information acquisition unit, the image technical information sub-acquisition unit and the image content characteristic sub-acquisition unit, the second acquisition module realizes the refinement of data acquisition, so that the method can deeply analyze the low-level quality characteristics of the image and provide a basis for high-precision image processing. These detailed information can effectively locate and resolve complex image problems, thereby improving the pertinence and accuracy of the analysis. The acquired parameters can complement the first multi-source data set to form a comprehensive multi-source data analysis system. By supplementing the subtle characteristic data that the first acquisition module fails to cover, such as high-frequency component intensity and interpolation error, this module further improves the method's adaptability to image diversity, allowing the enhancement algorithm to adapt to more complex scene requirements.

[0062] 3. The first analysis module calculates the clarity recognition coefficient QXD and compares it with the preset threshold value Y to draw a conclusion on whether image enhancement is needed. Through this judgment mechanism based on quantitative calculation, compared with the traditional enhancement decision-making that relies on manual experience, this method realizes automatic and accurate judgment of image processing needs. The start-up conditions of image enhancement are clear, which not only avoids unnecessary enhancement operations, but also ensures processing efficiency and improves the intelligence level of the method. The second analysis module can automatically select the most appropriate enhancement method according to the actual quality and needs of the image. This flexible image processing mechanism can adapt to the needs of different fields and scenes. Images of different qualities will receive the most appropriate enhancement processing, which not only saves computing resources, but also provides accurate and personalized optimization solutions according to the specific image conditions. By setting different hierarchical conditions in the enhancement strategy, the second analysis module avoids overly complex processing of all images. Each enhancement level selects the corresponding calculation method according to the severity of the image quality problem, ensuring efficient processing while reducing unnecessary computational complexity, thereby improving processing speed. BRIEF DESCRIPTION OF THE DRAWINGS

[0063] Figure 1 It is a diagram of the method steps of the present invention.

[0064] In the figure: 1. First acquisition module; 11. Image quality main information acquisition unit; 12. Image technical information main acquisition unit; 13. Image content characteristic main acquisition unit; 2. Second acquisition module; 21. Image quality sub-information acquisition unit; 22. Image technical information sub-acquisition unit; 23. Image content characteristic sub-acquisition unit; 3. Data processing module; 4. First calculation module; 5. Second calculation module; 6. First analysis module; 7. Second analysis module; 8. Execution module. DETAILED DESCRIPTION

[0065] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0066] Example 1: Please refer to Figure 1 A computer image analysis method based on big data includes a first acquisition module 1, a second acquisition module 2, a data processing module 3, a first calculation module 4, a second calculation module 5, a first analysis module 6, a second analysis module 7 and an execution module 8. The specific steps are as follows:

[0067] S1, through the first acquisition module 1, the main information of the current image is collected, and the relevant corresponding data is retrieved from the cloud platform to form a first multi-source data set; at the same time, the second acquisition module 2 collects the sub-information of the image and integrates it into a second multi-source data set;

[0068] S2, preprocessing and dimensionlessizing the first multi-source data set and the second multi-source data set through the data processing module 3 to ensure data standardization and facilitate subsequent calculations. The preprocessed data is further classified to generate a first data set, a second data set, a third data set, a fourth data set, a fifth data set and a sixth data set, covering information such as image quality, resolution, and texture complexity;

[0069] S3, integrating and calculating the first data set, the second data set and the third data set through the first calculation module 4 to generate a clarity recognition coefficient QXD;

[0070] S4, through the first analysis module 6, the generated clarity recognition coefficient QXD is compared with the preset first threshold Y to generate a first comparison result. If QXD is less than Y, it means that the current image needs to be enhanced, and the next step is entered; if QXD is greater than or equal to Y, it means that the current image does not need to be enhanced, and the method ends;

[0071] S5, if enhancement is required, the second calculation module 5 will generate an enhancement magnitude coefficient ZQL by integrating the fourth data set, the fifth data set and the sixth data set;

[0072] S6. Through the second analysis module 7, the enhancement magnitude coefficient ZQL is compared with the preset second threshold R to generate a second comparison result. According to the result, corresponding image enhancement operations are performed, such as low-intensity Laplace filtering, detail layer extraction, contrast adaptive equalization, etc. Finally, the corresponding enhancement strategy is applied through the execution module 8 to complete the image optimization.

[0073] In this embodiment: the first acquisition module 1 comprehensively acquires the main information of the image, retrieves the control data related to the current image from the cloud platform, and integrates it into the first multi-source data set. This module can obtain the core parameters of image quality, resolution and content characteristics, and provide accurate basic data for subsequent calculations. Relying on big data resources, the first acquisition module 1 improves the comprehensiveness and adaptability of the method to the main information of the current image, and provides a scientific basis for the design of image enhancement strategies.

[0074] The second acquisition module 2 focuses on the acquisition of image sub-information, including refined parameters such as edge characteristics, texture details and noise types, and integrates them into the second multi-source data set. This module makes up for the shortcomings of the single information acquisition method and provides more dimensional reference data for subsequent data processing and calculation. By accurately extracting the sub-characteristic information of the image, the second acquisition module 2 enhances the ability to capture complex image structures, laying the foundation for improving the flexibility of enhancement processing.

[0075] The data processing module 3 preprocesses and decomposes the first multi-source data set and the second multi-source data set, and organizes them into the first data set, the second data set, the third data set, the fourth data set, the fifth data set, and the sixth data set. Through dimensionless and standardized processing, the data processing module 3 improves the comparability and calculation efficiency of the data, and avoids the error accumulation caused by the difference in the magnitude of the original data. The scientific allocation of modules enables the subsequent calculation modules to focus on specific image characteristics, optimizing the calculation accuracy and execution efficiency.

[0076] The first calculation module 4 is responsible for integrating the first data set, the second data set and the third data set, and calculating the clarity recognition coefficient QXD to judge the quality level of the image. The first calculation module 4 realizes intelligent recognition of the current image clarity. Its precise calculation reduces the interference of human subjective judgment and significantly improves the scientificity and objectivity of the enhancement demand judgment.

[0077] The first analysis module 6 generates a first comparison result and determines whether the image needs to be enhanced by comparing the clarity recognition coefficient QXD with the first threshold value Y. The first analysis module 6 realizes the automatic judgment of the image enhancement requirement, ensuring that the image is optimized only when necessary, thereby improving the operation efficiency of the method and avoiding waste of resources.

[0078] The second calculation module 5 integrates and calculates the fourth data set, the fifth data set and the sixth data set based on the judgment result of the first analysis module 6 to generate the enhancement magnitude coefficient ZQL. The second calculation module 5 determines the enhancement level suitable for the current image. Its refined magnitude calculation provides precise guidance for the implementation of subsequent enhancement strategies, ensuring the balance and pertinence of the enhancement effect.

[0079] The second analysis module 7 generates a second comparison result by comparing the enhancement magnitude coefficient ZQL with the second threshold R. The second analysis module 7 accurately determines the strength level of the enhancement strategy and provides differentiated processing solutions for images of different qualities. The introduction of the module improves the method's ability to adapt to complex image characteristics and reduces the risk of distortion during the enhancement process.

[0080] The execution module 8 implements specific enhancement operations according to the results of the second analysis module 7, including different enhancement strategies from level one to level three. Through meticulous processing methods such as Laplace filtering, contrast equalization, and blind area convolution, the execution module 8 effectively improves the visual quality and information expression capabilities of the image. The execution module 8 takes into account both execution efficiency and processing effect, providing a feasible and efficient implementation path for image enhancement.

[0081] Compared with traditional technologies, this method achieves multi-source data drive, dynamic threshold determination and hierarchical enhancement strategy through the collaborative work of multiple modules, avoiding the problems of over-enhancement and under-enhancement, and significantly improving the reliability and intelligence level of image enhancement. This method has strong adaptability in different scenarios, optimizes the accuracy and efficiency of image quality processing, and provides a scientific and efficient solution for the application of computer vision technology.

[0082] Example 2: Please refer to Figure 1 , the first acquisition module 1 includes an image quality main information acquisition unit 11, an image technical information main acquisition unit 12 and an image content characteristic main acquisition unit 13;

[0083] The image quality main information acquisition unit 11 is used to acquire noise intensity, contrast range and edge clarity;

[0084] The image technical information main acquisition unit 12 is used to acquire image resolution, sampling rate and zoom ratio;

[0085] The image content characteristic main acquisition unit 13 is used to acquire texture complexity, illumination uniformity and dynamic range;

[0086] Noise intensity, contrast range, edge sharpness, image resolution, sampling rate, scaling ratio, texture complexity, illumination uniformity, and dynamic range constitute the first multi-source dataset.

[0087] In this embodiment: the first acquisition module 1 comprehensively acquires the main information of the image from multiple dimensions by introducing the image quality main information acquisition unit 11, the image technical information main acquisition unit 12 and the image content characteristic main acquisition unit 13. This multi-dimensional acquisition breaks through the limitations of traditional single indicator acquisition and more comprehensively reflects the overall characteristics of the image. By integrating the noise intensity, contrast range, edge clarity, image resolution, sampling rate, scaling ratio, texture complexity, illumination uniformity and dynamic range into the first multi-source data set, the data is not only structured, but also efficiently analyzed and utilized. The module design incorporates a variety of key image characteristics into the acquisition scope, has a wide range of applicability, can adapt to different types of image scenes and requirements, and ensures the flexibility and adaptability of the method processing.

[0088] The multi-unit design ensures that the collected information covers the core indicators of image quality, technical characteristics and content characteristics, providing a reliable data basis for subsequent image analysis and enhancement. By integrating the first multi-source data set, the method avoids the errors caused by data one-sidedness and insufficiency, and provides more scientific data support for image clarity identification and enhancement demand determination. The structured integration of the first acquisition module 1 reduces the complexity of the data processing process, assigns clear task objectives to subsequent calculation modules, and improves the overall computing efficiency and intelligence level of the method. In a big data-driven environment, the first acquisition module 1 can quickly collect and integrate control data, and combine it with real-time image information, with dynamic adjustment and rapid response capabilities, further enhancing the practicality and scalability of the method in actual scenarios.

[0089] Example 3: Please refer to Figure 1 , the second acquisition module 2 includes an image quality sub-information acquisition unit 21, an image technical information sub-acquisition unit 22 and an image content characteristic sub-acquisition unit 23;

[0090] The image quality sub-information acquisition unit 21 is used to acquire noise variance, Gaussian noise ratio and signal-to-noise ratio;

[0091] The image technical information sub-collection unit 22 is used to collect pixel spacing, sampling depth and interpolation error;

[0092] The image content characteristic sub-collection unit 23 is used to collect the high-frequency component intensity, edge pixel ratio and local illumination deviation structure;

[0093] Noise variance, Gaussian noise proportion, signal-to-noise ratio, pixel spacing, sampling depth, interpolation error, high-frequency component intensity, edge pixel ratio, and local illumination deviation constitute the second multi-source dataset.

[0094] In this embodiment: by setting up the image quality sub-information acquisition unit 21, the image technical information sub-acquisition unit 22 and the image content characteristic sub-acquisition unit 23, the second acquisition module 2 realizes the refinement of data acquisition, so that the method can deeply analyze the low-level quality characteristics of the image, providing a basis for high-precision image processing. These detailed information can effectively locate and analyze complex image problems, thereby improving the pertinence and accuracy of the analysis. The acquired parameters can complement the first multi-source data set to form a comprehensive multi-source data analysis system. By supplementing the subtle characteristic data that the first acquisition module 1 fails to cover, such as high-frequency component intensity and interpolation error, this module further improves the method's adaptability to image diversity, allowing the enhancement algorithm to adapt to more complex scene requirements.

[0095] The collected sub-information can quickly capture the changing characteristics of image details. This detail capture capability enables the method to have higher response efficiency and processing capabilities for dynamic image problems in a real-time environment. By providing more refined sub-information, it provides high-quality input data for the follow-up, making the judgment of the enhancement level more accurate, avoiding over-enhancement or under-enhancement due to insufficient basic data. In the context of big data, it can quickly and efficiently collect complex and diverse image detail data.

[0096] Example 4: Please refer to Figure 1 The data processing module 3 is used to preprocess and dimensionlessly transform the first multi-source data set and the second multi-source data set, and divide the processed first multi-source data set and the second multi-source data set into a first data set, a second data set, a third data set, a fourth data set, a fifth data set and a sixth data set;

[0097] The first data set includes noise intensity, contrast range, and edge definition;

[0098] The noise intensity is divided into actual image noise intensity A1 and required noise intensity A2;

[0099] The contrast range is divided into the actual contrast range B1 and the required contrast range B2;

[0100] Edge clarity is divided into actual edge clarity C1 and required edge clarity C2;

[0101] The second data set includes image resolution, sampling rate, and scaling ratio;

[0102] Image resolution is divided into actual image resolution D1 and required image resolution D2;

[0103] The sampling rate is divided into actual sampling rate E1 and required sampling rate E2;

[0104] The scaling ratio is divided into the actual scaling ratio F1 and the required scaling ratio F2;

[0105] The third dataset includes texture complexity, lighting uniformity, and dynamic range;

[0106] Texture complexity is divided into actual texture complexity G1 and required texture complexity G2;

[0107] Light uniformity is divided into actual light uniformity H1 and required light uniformity H2;

[0108] The dynamic range structure is divided into the actual dynamic range structure I1 and the required dynamic range structure I2;

[0109] The fourth data set includes noise variance J, Gaussian noise weight K, and signal-to-noise ratio L;

[0110] The fifth data set includes a pixel spacing M, a sampling depth N, and an interpolation error O;

[0111] The sixth data set includes high-frequency component intensity P, edge pixel ratio Q, and local illumination deviation S.

[0112] In this embodiment: by preprocessing and dimensionless the first multi-source data set and the second multi-source data set, the data processing module 3 effectively eliminates the differences between different parameter units and dimensions. This processing method not only reduces the complexity of subsequent analysis and calculation, but also improves the comparability between data, ensures the accuracy and consistency of the analysis results, and divides the processed multi-source data set into six sub-data sets, so that each set of data can focus on the analysis needs of a specific field. This fine grouping facilitates the rapid location and resolution of targeted problems, thereby enhancing the adaptability and accuracy of the method.

[0113] The data division within the data processing module 3 module clarifies the correspondence between the actual value and the required value, providing accurate difference data support for the subsequent calculation module. This demand-driven analysis method can quickly identify the gap between the current image and the ideal state, thereby providing a scientific basis for the formulation of the enhancement level and specific optimization strategy. Through the refined processing of the fourth, fifth and sixth data sets, the method can optimize the noise, resolution and detail texture respectively, accurately locate the detail blurred area and formulate a recovery strategy.

[0114] The six data sets cover information of each key dimension in the image processing process, providing comprehensive and accurate input data for the intelligent enhancement algorithm of the subsequent modules. Thanks to the standardized operation of data processing module 3, the method can reduce unnecessary redundant calculations in the analysis process and significantly improve the operation efficiency.

[0115] Example 5: Please refer to Figure 1 The first calculation module 4 calculates and obtains the clarity recognition coefficient QXD by the following formula:

[0116] QXD=a1×Z1+a2×Z2+a3×Z3;

[0117] Wherein: Z1 is the first main reference coefficient, which is calculated by integrating the first data set, Z2 is the second main reference coefficient, which is calculated by integrating the second data set, Z3 is the third main reference coefficient, which is calculated by integrating the third data set, a1, a2 and a3 are weight values, and the values ​​of a1, a2 and a3 are adjusted and set by the user.

[0118] The first calculation module 4 calculates and obtains the first main reference coefficient Z1, the second main reference coefficient Z2 and the third main reference coefficient Z3 by the following formulas:

[0119]

[0120] Wherein: A1, B1, C1, D1, E1, F1, G1, H1 and I1 are the actual noise intensity, actual contrast range, actual edge definition, actual image resolution, actual sampling rate, actual scaling ratio, actual texture complexity, actual illumination uniformity and actual dynamic range respectively;

[0121] A2, B2, C2, D2, E2, F2, G2, H2 and I2 are the required noise intensity, required contrast range, required edge clarity, required image resolution, required sampling rate, required scaling ratio, required texture complexity, required illumination uniformity and required dynamic range, respectively;

[0122] c1, c2, c3 are weight values, and the values ​​of c1, c2 and c3 are adjusted and set by the user.

[0123] In this embodiment: a comprehensive quantitative evaluation of image quality is achieved. The three main reference coefficients correspond to image characteristics of different dimensions, and the weights can be flexibly adjusted by the user, which significantly enhances the adaptability of the method to a variety of application scenarios. Compared with the traditional method that relies on a single parameter, this method has achieved breakthroughs in both accuracy and flexibility, and users can adjust it according to actual needs. This flexibility supports a variety of application scenarios. For example, for high-precision scenes, the weight of edge clarity can be increased preferentially, and for image compression, the weight of texture complexity can be increased, thereby achieving customized analysis.

[0124] By comparing the difference between the actual value and the required value, the improvement space of the image can be accurately located, and the improvement target of the scene illumination can be quantified. This demand-driven approach significantly improves the scientific nature of the evaluation and the guiding significance of the results.

[0125] Through reasonable data grouping and formula optimization, the first, second and third data sets are integrated for calculation, avoiding repeated calls or redundant analysis. In big data scenarios, this method can effectively improve computing efficiency, reduce hardware resource usage, and ensure the accuracy of calculation results.

[0126] The introduction of the clarity recognition coefficient QXD provides a clear numerical basis for the subsequent judgment of whether the current image needs to be enhanced. This method can quickly and accurately judge whether to perform subsequent enhancement operations, thereby avoiding unnecessary processing.

[0127] The module can make reasonable adjustments for different scenes, such as low-light environments and high dynamic range scenes, and adapt to a variety of complex image characteristics by dynamically adjusting weights and required value parameters. Especially in fields with high precision requirements such as medical imaging and remote sensing imaging, it can significantly improve the reliability and applicability of clarity assessment.

[0128] The design of the first calculation module 4 integrates multi-dimensional characteristics into a unified evaluation index through an innovative calculation formula for the clarity recognition coefficient, overcoming the problems of single indicators and incomplete evaluation in traditional image analysis methods. At the same time, the introduction of a flexible weight adjustment mechanism and demand-driven analysis further enhances the scene adaptability and optimization efficiency of the method. Overall, this module not only improves the accuracy of image clarity assessment, but also provides a scientific basis for subsequent enhancement operations, effectively promoting the development of image processing technology towards intelligence and multifunctionality.

[0129] Example 6: Please refer to Figure 1 , the second calculation module 5 calculates and obtains the enhancement magnitude coefficient ZQL through the following formula:

[0130] ZQL=b1×F1+b2×F2+b3×F3;

[0131] In the formula: F1 is the first sub-reference coefficient, which is calculated by integrating the fourth data set, F2 is the second sub-reference coefficient, which is calculated by integrating the fifth data set, F3 is the third sub-reference coefficient, which is calculated by integrating the sixth data set, b1, b2 and b3 are weight values, and the values ​​of b1, b2 and b3 are adjusted and set by the user.

[0132] The second calculation module 5 calculates and obtains the first sub-reference coefficient F1, the second sub-reference coefficient F2 and the third sub-reference coefficient F3 by the following formulas:

[0133]

[0134]

[0135] Where: L is the signal-to-noise ratio, J is the noise variance, K is the Gaussian noise ratio, M is the pixel spacing, N is the sampling depth, O is the interpolation error, P is the high-frequency component intensity, Q is the edge pixel ratio, and S is the local illumination deviation.

[0136] In this embodiment: the second calculation module 5 integrates multiple sub-reference coefficients by calculating the enhancement magnitude coefficient ZQL, ensuring the quantitative control of the image enhancement operation. Compared with the simple enhancement scheme in the traditional method, this method makes the enhancement effect more in line with the actual needs of the image through the sophisticated sub-reference coefficient calculation, thereby improving the accuracy of image enhancement and the controllability of the effect.

[0137] By introducing user-defined weights, the second calculation module 5 enhances the flexibility and personalized customization ability of image enhancement. Users can adjust the weights according to different application scenarios such as medical imaging, industrial image analysis, etc., so as to obtain the best enhancement effect in different fields. This flexible adjustment mechanism greatly expands the applicable range of the method, avoids the limitation that traditional fixed algorithms cannot meet special requirements, can complete image enhancement processing in a short time, ensures the high performance and fast response of the method. By comprehensively calculating multiple factors affecting image quality, the second calculation module 5 can quantitatively evaluate the image quality from multiple perspectives, making the enhancement decision more scientific and reasonable, avoiding unnecessary processing of high-quality images, and at the same time enhancing the visual effect of low-quality images.

[0138] Example 7: Please refer to Figure 1 , the first comparison result generated by the first analysis module 6 is as follows:

[0139] When QXD < Y, it means that the current image needs to be enhanced. When QXD ≥ Y, it means that the current image does not need to be enhanced.

[0140] The specific second comparison result generated by the second analysis module 7 is as follows:

[0141] When R × 95% < ZQL ≤ R, it means that the current picture needs to be enhanced at the first level. Apply a low-intensity Laplace filter to the current picture and enhance the edge sharpness. According to the bright and dark areas of the picture, locally increase the brightness by about 5% - 10%, and use a median filter to slightly denoise to prevent noise diffusion caused by enhanced edges;

[0142] When R × 85% < ZQL ≤ R × 95%, it means that the current picture needs to be enhanced at the second level. Use a multi-scale Gaussian pyramid to extract the detail layer of the image for the current picture, enhance its texture features, based on the contrast-limited adaptive histogram equalization algorithm, enhance the local contrast, and uniformize the brightness distribution. Combine with a bilateral filter to reduce the interference of noise on edge details;

[0143] When ZQL ≤ R × 85%, it means that the current picture needs to be enhanced at the third level. Use a blind deconvolution algorithm to estimate the blur kernel and restore the current picture, globally apply Gamma correction to enhance the brightness and contrast, and use non-local means denoising to significantly reduce noise interference.

[0144] In this embodiment: the first analysis module 6 calculates the clarity recognition coefficient QXD and compares it with the preset threshold value Y to draw a conclusion on whether image enhancement is needed. Through this judgment mechanism based on quantitative calculation, compared with the traditional enhancement decision-making that relies on manual experience, this method realizes automatic and accurate image processing demand judgment. The starting conditions of image enhancement are clear, which not only avoids unnecessary enhancement operations, but also ensures processing efficiency and improves the intelligence level of the method.

[0145] The second analysis module 7 provides a sophisticated enhancement solution for different image quality issues by setting a multi-level comparison judgment according to the calculation result of the enhancement magnitude coefficient ZQL:

[0146] First-level enhancement: For minor image quality issues, such as blurred edges or slight brightness unevenness, a low-intensity Laplacian filter and brightness adjustment are used, combined with mild denoising. This strategy can quickly and gently optimize the image, improve the clarity of details, while avoiding over-enhancement.

[0147] Secondary enhancement: When image quality issues are more obvious, a multi-scale Gaussian pyramid is used to extract detail layers, combined with a contrast-limited adaptive histogram equalization algorithm to improve local contrast, enhance texture details, and use a bilateral filter to reduce noise interference on edges. This hierarchical enhancement method ensures that both the overall image and details can be effectively processed, while taking into account noise control and avoiding detail loss.

[0148] Level 3 enhancement: For images with severe blur or loss of details, the blind convolution algorithm is used for blur kernel estimation and image restoration. At the same time, the global gamma correction is used to enhance brightness and contrast, and the non-local mean denoising algorithm is combined for noise suppression. This method ensures that even in a high-noise environment, the details and clarity of the image can still be effectively restored.

[0149] This hierarchical strategy enables each enhancement method to solve specific image quality problems in a targeted manner. Different images receive different intensities and types of processing according to their needs, avoiding over-enhancement or under-enhancement, and improving the personalization and accuracy of processing quality and effects.

[0150] Through different levels of enhancement schemes, the second analysis module 7 can automatically select the most appropriate enhancement method according to the actual quality and needs of the image. This flexible image processing mechanism can adapt to the needs of different fields and scenes. Images of different qualities will receive the most appropriate enhancement processing, which not only saves computing resources, but also provides accurate and personalized optimization solutions based on the specific image conditions. By setting different hierarchical conditions in the enhancement strategy, the second analysis module 7 avoids overly complex processing of all images. Each enhancement level selects the corresponding calculation method according to the severity of the image quality problem, ensuring efficient processing while reducing unnecessary calculation complexity, thereby improving processing speed.

[0151] By introducing precise contrast judgment and hierarchical enhancement strategies through the first analysis module 6 and the second analysis module 7, the entire image processing process becomes more intelligent and personalized. Compared with the traditional image enhancement method, the automatic judgment based on the clarity recognition coefficient and the enhancement magnitude coefficient makes the processing process more scientific, avoids unnecessary enhancement operations, and ensures the targeted processing of images of different qualities. The hierarchical enhancement scheme allows different types of images to be processed appropriately, thereby significantly improving the visual effect and detail clarity of the image, and further optimizing the application effect of the image enhancement technology.

[0152] The contents not described in detail in this specification belong to the prior art known to professional and technical personnel in this field.

[0153] Although the present invention has been described in detail with reference to the aforementioned embodiments, it is still possible for those skilled in the art to modify the technical solutions described in the aforementioned embodiments, or to make equivalent substitutions for some of the technical features therein. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A computer image analysis method based on big data, characterized in that: The method comprises a first acquisition module (1), a second acquisition module (2), a data processing module (3), a first calculation module (4), a second calculation module (5), a first analysis module (6), a second analysis module (7) and an execution module (8), wherein the specific steps are as follows: S1. The main information of the current image is collected through the first acquisition module (1), the main information including noise intensity, contrast range, edge clarity, image resolution, sampling rate, scaling ratio, texture complexity, illumination uniformity and dynamic range, and relevant corresponding data are retrieved from the cloud platform and integrated to form a first multi-source data set; at the same time, the second acquisition module (2) collects sub-information of the image, the sub-information including noise variance, Gaussian noise ratio, signal-to-noise ratio, pixel spacing, sampling depth, interpolation error, high-frequency component intensity, edge pixel ratio and local illumination deviation structure, and integrates them into a second multi-source data set; S2, preprocessing and dimensionlessizing the first multi-source data set and the second multi-source data set through a data processing module (3) to ensure data standardization and facilitate subsequent calculations, and the preprocessed data is further classified to generate a first data set, a second data set, a third data set, a fourth data set, a fifth data set and a sixth data set; The first data set includes noise intensity, contrast range, and edge definition; The second data set includes image resolution, sampling rate, and scaling ratio; The third dataset includes texture complexity, lighting uniformity, and dynamic range; The fourth data set includes noise variance J, Gaussian noise weight K, and signal-to-noise ratio L; The fifth data set includes a pixel spacing M, a sampling depth N, and an interpolation error O; The sixth data set includes the high-frequency component intensity P, the edge pixel ratio Q, and the local illumination deviation S; S3, integrating and calculating the first data set, the second data set and the third data set through the first calculation module (4) to generate a clarity recognition coefficient QXD; S4, through the first analysis module (6), the generated clarity recognition coefficient QXD is compared with a preset first threshold value Y to generate a first comparison result. If QXD is less than Y, it means that the current image needs to be enhanced, and the next step is entered; if QXD is greater than or equal to Y, it means that the current image does not need to be enhanced, and the method ends; S5, if enhancement is required, the second calculation module (5) will perform calculations by integrating the fourth data set, the fifth data set and the sixth data set to generate an enhancement magnitude coefficient ZQL; S6. The enhancement magnitude coefficient ZQL is compared with a preset second threshold R through the second analysis module (7), a second comparison result is generated, and a corresponding image enhancement level is determined. Finally, the corresponding enhancement strategy is applied through the execution module (8) to complete the image optimization.

2. The computer image analysis method based on big data according to claim 1, characterized in that: The first acquisition module (1) comprises an image quality main information acquisition unit (11), an image technical information main acquisition unit (12) and an image content characteristic main acquisition unit (13); The image quality main information acquisition unit (11) is used to acquire noise intensity, contrast range and edge clarity; The image technical information main acquisition unit (12) is used to acquire image resolution, sampling rate and zoom ratio; The image content characteristic main acquisition unit (13) is used to acquire texture complexity, illumination uniformity and dynamic range; The noise intensity, contrast range, edge definition, image resolution, sampling rate, scaling ratio, texture complexity, illumination uniformity, and dynamic range constitute a first multi-source data set.

3. The computer image analysis method based on big data according to claim 2, characterized in that: The second acquisition module (2) comprises an image quality sub-information acquisition unit (21), an image technical information sub-acquisition unit (22) and an image content characteristic sub-acquisition unit (23); The image quality sub-information acquisition unit (21) is used to acquire noise variance, Gaussian noise ratio and signal-to-noise ratio; The image technical information sub-collection unit (22) is used to collect pixel spacing, sampling depth and interpolation error; The image content characteristic sub-collection unit (23) is used to collect the intensity of high-frequency components, the ratio of edge pixels and the local illumination deviation; The noise variance, Gaussian noise proportion, signal-to-noise ratio, pixel spacing, sampling depth, interpolation error, high-frequency component intensity, edge pixel ratio and local illumination deviation constitute a second multi-source data set.

4. The computer image analysis method based on big data according to claim 3, characterized in that: The data processing module (3) is used to preprocess and dimensionlessly process the first multi-source data set and the second multi-source data set, and divide the processed first multi-source data set and the second multi-source data set into a first data set, a second data set, a third data set, a fourth data set, a fifth data set and a sixth data set; The first data set includes noise intensity, contrast range, and edge definition; The noise intensity is divided into actual image noise intensity A1 and required noise intensity A2; The contrast range is divided into the actual contrast range B1 and the required contrast range B2; Edge clarity is divided into actual edge clarity C1 and required edge clarity C2; The second data set includes image resolution, sampling rate, and scaling ratio; Image resolution is divided into actual image resolution D1 and required image resolution D2; The sampling rate is divided into actual sampling rate E1 and required sampling rate E2; The scaling ratio is divided into the actual scaling ratio F1 and the required scaling ratio F2; The third dataset includes texture complexity, lighting uniformity, and dynamic range; Texture complexity is divided into actual texture complexity G1 and required texture complexity G2; Light uniformity is divided into actual light uniformity H1 and required light uniformity H2; The dynamic range structure is divided into the actual dynamic range structure I1 and the required dynamic range structure I2; The fourth data set includes noise variance J, Gaussian noise weight K, and signal-to-noise ratio L; The fifth data set includes a pixel spacing M, a sampling depth N, and an interpolation error O; The sixth data set includes high-frequency component intensity P, edge pixel ratio Q, and local illumination deviation S.

5. The computer image analysis method based on big data according to claim 4, characterized in that: The first calculation module (4) calculates and obtains the clarity recognition coefficient QXD by the following formula: QXD=a1×Z1+a2×Z2+a3×Z3; Where: Z1 is the first main reference coefficient, obtained by integrating the first data set; Z2 is the second main reference coefficient, obtained by integrating the second data set; Z3 is the third main reference coefficient, obtained by integrating the third data set; a1, a2, and a3 are weight values, and the values of a1, a2, and a3 are adjusted and set by the user.

6. The computer image analysis method based on big data according to claim 5, characterized in that: The first calculation module (4) calculates and obtains the first main reference coefficient Z1, the second main reference coefficient Z2, and the third main reference coefficient Z3 respectively through the following formulas: Where: A1, B1, C1, D1, E1, F1, G1, H1, and I1 are the actual noise intensity, actual contrast range, actual edge sharpness, actual image resolution, actual sampling rate, actual scaling ratio, actual texture complexity, actual illumination uniformity, and actual dynamic range respectively; A2, B2, C2, D2, E2, F2, G2, H2, and I2 are the required noise intensity, required contrast range, required edge sharpness, required image resolution, required sampling rate, required scaling ratio, required texture complexity, required illumination uniformity, and required dynamic range respectively; c1, c2, and c3 are weight values, and the values of c1, c2, and c3 are adjusted and set by the user.

7. The computer image analysis method based on big data according to claim 6, characterized in that: The second calculation module (5) calculates and obtains the enhancement magnitude coefficient ZQL through the following formula: ZQL = b1×F1 + b2×F2 + b3×F3; Where: F1 is the first sub-reference coefficient, obtained by integrating the fourth data set; F2 is the second sub-reference coefficient, obtained by integrating the fifth data set; F3 is the third sub-reference coefficient, obtained by integrating the sixth data set; b1, b2, and b3 are weight values, and the values of b1, b2, and b3 are adjusted and set by the user.

8. The computer image analysis method based on big data according to claim 7, characterized in that: The second calculation module (5) calculates and obtains the first sub-reference coefficient F1, the second sub-reference coefficient F2, and the third sub-reference coefficient F3 respectively through the following formulas: Where: L is the signal-to-noise ratio, J is the noise variance, K is the Gaussian noise ratio, M is the pixel pitch, N is the sampling depth, O is the interpolation error, P is the high-frequency component intensity, Q is the edge pixel ratio, and S is the local illumination deviation.

9. The computer image analysis method based on big data according to claim 8, characterized in that: The first comparison result generated by the first analysis module (6) is as follows: When QXD < Y, it means that the current image needs to be enhanced; when QXD ≥ Y, it means that the current image does not need to be enhanced.

10. The computer image analysis method based on big data according to claim 9, characterized in that: The second comparison result generated by the second analysis module (7) is specifically as follows: When R×95% < ZQL ≤ R, it means that the current picture needs to be enhanced at the first level. Apply a low-intensity Laplace filter to the current picture, and enhance the edge sharpness. According to the bright and dark areas of the picture, locally increase the brightness by about 5% - 10%, and use a median filter to slightly denoise to prevent noise diffusion caused by enhanced edges; When R×85% < ZQL ≤ R×95%, it represents that the current image needs secondary enhancement. The detail layer of the image is extracted using a multi-scale Gaussian pyramid for the current image to enhance its texture features. Based on the contrast-limited adaptive histogram equalization algorithm, the local contrast is enhanced, and the brightness distribution is homogenized. Combining with a bilateral filter, the interference of noise on edge details is reduced; When ZQL ≤ R×85%, it represents that the current image needs tertiary enhancement. The blind deconvolution algorithm is used to estimate the blur kernel and restore the current image. Gamma correction is globally applied to enhance brightness and contrast, and non-local means denoising is used to significantly reduce noise interference.

Images