Urinary system image adaptive feature optimization and enhancement method
By employing multi-scale fusion feature extraction and adaptive denoising techniques, the problem of insufficient image detail preservation in traditional methods is solved, achieving efficient and accurate processing of digital images in urology and improving diagnostic accuracy.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- QILU HOSPITAL(QINGDAO) CHEELOO COLLEGE OF MEDICINE SHANDONG UNIV
- Filing Date
- 2026-02-03
- Publication Date
- 2026-05-15
AI Technical Summary
Traditional digital image processing methods in urology struggle to preserve image details in areas with low contrast and high noise levels, and their processing speed is slow, failing to meet the needs of real-time clinical processing. Existing deep learning-based methods also have shortcomings in noise removal and detail preservation.
By employing multi-scale fusion feature extraction, local adaptive weighted denoising, and dual-domain feature dynamic fusion techniques, and by adapting the local window size and noise weighting factor, we can accurately process regions with different brightness levels, thereby achieving noise removal and detail enhancement.
It significantly improves image quality and diagnostic accuracy, has stronger adaptability and real-time processing capabilities, and meets the clinical demand for efficient and high-quality image analysis.
Smart Images
Figure CN122048696A_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of image enhancement, and specifically relates to an adaptive feature optimization and enhancement method for urinary system images. Background Technology
[0002] Traditional digital image processing methods in urology mainly rely on basic image enhancement and denoising techniques, such as histogram equalization, filtering algorithms, and traditional noise suppression methods. These methods often fail to fully preserve the details in the image, especially in areas with low contrast and high noise, which can easily lead to the loss of key information. In addition, traditional methods are slow and difficult to meet the needs of real-time processing in clinical practice.
[0003] Currently, although some deep learning-based image processing methods have been proposed, the processing of bright and dark areas and the enhancement of weak tissue structures in urological digital images still face challenges. These methods are insufficient in balancing noise removal and detail preservation, especially when the image quality is low, noise and important information are difficult to separate effectively. Therefore, there is still considerable room for improvement in existing methods, especially in clinical applications where there is a need for efficient and accurate image processing.
[0004] The intelligent digital image processing method for urology proposed in this invention effectively overcomes the shortcomings of traditional methods through multi-scale fusion feature extraction, local adaptive weighted denoising, and dual-domain feature dynamic fusion technology. This method can accurately process regions with different brightness levels, achieve refined noise removal and detail enhancement, and significantly improve image quality and diagnostic accuracy through intelligent fusion of brightness and darkness features. It has stronger adaptability and real-time processing capabilities, meeting the needs of efficient and high-quality image analysis in clinical practice. Summary of the Invention
[0005] This invention provides an adaptive feature optimization and enhancement method for urinary system images, which aims to improve the detail representation and diagnostic accuracy of digital images in urology through multi-level image enhancement and denoising techniques.
[0006] This invention aims to propose an intelligent digital image enhancement model for urology, and to provide a method for adaptive feature optimization and enhancement of urinary system images, including the following steps: S1. Acquire original X-ray images of the urinary system, preprocess the original images using a scanner, and construct a digital image dataset of the urinary system; S2. Based on the adaptive local window size, an adaptive multi-scale weighting coefficient is designed, and a local statistical driving feature module is constructed. The intelligent digital image of urology generates multi-scale fusion features through the local statistical driving feature module. S3. Design and implement an adaptive noise weighting factor, construct a region weighted separation and denoising module, and obtain bright and dark features based on mask reconstruction by passing the region weighted separation and denoising module through the multi-scale fusion features. S4. By using adaptive gradient entropy to fuse weights, a dual-domain feature dynamic fusion module is constructed. Bright and dark features are generated through the dual-domain feature dynamic fusion module to produce an enhanced intelligent digital image for urology. S5 integrates a local statistical driven feature module, a region weighted separation and denoising module, and a dual-domain feature dynamic fusion module to construct an intelligent digital image enhancement model for urology. S6. Train the intelligent digital image enhancement model for urology. Input the original intelligent digital image for urology into the trained intelligent digital image enhancement model for enhancement, and output a high-resolution intelligent digital image for urology.
[0007] Preferably, in S1, constructing a urological intelligent digital image dataset specifically includes the following steps: using an ultrasound medical imaging device to take pictures of a selected area, acquiring images according to clinical needs, obtaining raw image data of three organs including the kidney, bladder, and prostate, converting the analog images into digital format using a scanner, and specifically cropping the acquired raw images to a standard size of 320×320 pixels.
[0008] Preferably, low-resolution intelligent digital images of urology often exhibit problems such as unclear lesion boundaries, easy obscuring of minute pathological features, uneven gray-scale distribution of local tissues, and increased impact of noise on the overall structure. Conventional algorithms struggle to distinguish between important anatomical structures and weak lesion signals. Therefore, a local statistical-driven feature module is urgently needed to accurately extract and fuse details, textures, and structural features of different regions of the image, thereby improving the ability to express key information and suppressing noise interference.
[0009] Preferably, in S2, constructing the local statistically driven feature module specifically includes the following steps: Step S21: Design an adaptive local window size. The mathematical model is as follows: ; in, To adapt to the local window size, The parameters are learnable and updated through backpropagation. To normalize the activation function, compress the output to... , Original intelligent digital images for urology.
[0010] Step S22: Statistically analyze the values within the local window and calculate the local mean. The mathematical model is as follows: ; in, It is a local mean. This is represented as the location within this local window. The pixels; the local variance is calculated based on the local mean, and the mathematical model is: ; in, For the local variance, the local entropy is then calculated. The mathematical model is as follows: ; in, For local entropy, This represents the normalized probability distribution of pixel values within a local window. Based on local mean, local variance, and local entropy, an adaptive multi-scale weighted coefficient is designed and generated. The mathematical model is as follows: ; in, Represents the adaptive multi-scale weighted coefficients. and These are all adjustable parameters used to adjust the effects of local variance and local entropy. It is a Gaussian scale.
[0011] Step S23: Based on the adaptive multi-scale weighting coefficients, obtain the multi-scale fusion features. The mathematical model is as follows: ; in The standard deviation is Gaussian kernel, For convolution operations, Original intelligent digital images for urology, This is a multi-scale fusion feature.
[0012] Preferably, the advantage of step S21 in the local statistical driven feature module constructed through S2 is that it dynamically adjusts the pixel-level analysis region by adaptively adjusting the local window size, enabling feature extraction to flexibly adapt to different structures and complex details in medical images. The advantage of step S22 is that it automatically generates multi-scale weighting coefficients using multi-dimensional statistics such as local mean, variance, and entropy, realizing regional adaptive weight allocation when fusing features at different scales, further improving the accuracy and discriminative ability of feature expression. The advantage of step S23 is that the weighted integration of multi-scale features can fully integrate global and local statistical information, effectively enhancing the integrity and robustness of medical image feature expression. In summary, this module can realize intelligent adaptive fusion of multi-scale features in medical digital images, providing a more accurate and robust feature foundation for intelligent digital image enhancement in urology.
[0013] Preferably, multi-scale fusion features effectively preserve image details and global structure by combining information from different scales, especially in the presentation of details and subtle changes. However, the fusion of information from different scales may introduce noise, particularly in high-frequency detail areas. This noise may affect image quality and interfere with subsequent processing. Therefore, the next module designs a region-weighted separation denoising module. By using an adaptive noise weighting factor, the denoising intensity is flexibly adjusted according to the local standard deviation to accurately remove noise from different regions while maintaining the structural information of the image. In addition, the adaptive calculation of the brightness threshold and the generation of the region mask can effectively distinguish between bright and dark regions, thereby providing fine processing for feature reconstruction and ultimately improving image quality and detail performance.
[0014] Preferably, in S3, constructing the region-weighted separation and denoising module specifically includes the following steps: Step S31: Design and implement the adaptive noise weighting factor. The mathematical model is as follows: ; in, For window size Inner multi-scale fusion features standard deviation This is an adaptive noise weighting factor. The noise reduction factor controls the intensity of noise removal. Annoying optimization of multi-scale fusion features is performed based on an adaptive noise weighting factor; the mathematical model is as follows: ; in, To optimize the output structural composition image, For gradient magnitude, The amplitude is used; texture components are calculated based on the optimized output structural component image, and the mathematical model is as follows: ; in, These are the texture detail features obtained after difference.
[0015] Step S32: Calculate the brightness threshold. The mathematical model is as follows: ; in, The brightness threshold. The mean of the entire structural component image is given. The standard deviation of the entire structural component image is given. The adjustment coefficient is used to calculate the region mask; the mathematical model is as follows: ; ; in, For masking dark areas, This is a mask for the bright areas, allocated on a pixel-by-pixel basis.
[0016] Step S33: The bright and dark areas are reconstructed by weighted fusion. The mathematical model is as follows: ; ; in, For the bright part features after reconstruction based on the mask, This represents the dark area features after reconstruction based on the mask.
[0017] Preferably, the region-weighted separation denoising module constructed through S3 has the following advantages: Step S31 introduces an adaptive noise weighting factor to flexibly adjust the denoising intensity based on the local standard deviation of the multi-scale fusion features, achieving refined suppression of regions with different noise levels, improving the ability to preserve structural information and the adaptability of noise processing. The separation of structural components and texture components helps to highlight the expression of details. Step S32 adaptively calculates the brightness threshold and region mask based on the structural component image, which can accurately distinguish between bright and dark regions, providing a reliable basis for subsequent region feature reconstruction. Step S33 utilizes the region mask to achieve weighted fusion reconstruction of bright and dark features, making the expression of each region more consistent with the actual grayscale distribution, improving the restoration and representation of local bright and dark features of medical images. In summary, this module can effectively improve the denoising adaptability, detail expression, and bright and dark region feature reconstruction effect of intelligent digital images in urology.
[0018] Preferably, brightness features and shadow features exhibit different characteristics in an image. Brightness features primarily highlight details in bright areas, typically exhibiting high contrast and rich detail, but are susceptible to noise. Shadow features, on the other hand, focus on low-brightness areas, generally exhibiting less detail and making noise harder to detect, but are crucial for overall structure and depth. The next module generates adaptive fusion weights by calculating the local entropy and feature gradients of brightness and shadow features. This intelligently balances the processing of bright and dark areas, improving structural coherence and detail preservation. In morphological operations, brightness-guided shadow enhancement and shadow-guided brightness enhancement can precisely extract and strengthen weak tissue structures, further improving detail representation. Ultimately, the dynamic fusion of brightness, shadow, and mutually enhancing features ensures a natural transition in the image and a significant enhancement of pathological features, thereby improving the clinical readability and robustness of detail enhancement.
[0019] Preferably, in S4, constructing the dual-domain feature dynamic fusion module specifically includes the following steps: Step S41: Calculate the highlights features and dark features The characteristic gradient, mathematical model of which is: ; in, Represents the feature gradient. and These represent the gradients in the horizontal and vertical directions, respectively. Calculate highlight features and dark features The local entropy, mathematical model is: ; in, This represents the local entropy of the bright / dark regions. Represents grayscale value The probability distribution of grayscale values, ,in Indicates the local window radius. This indicates the number of pixels within a local window. Represents a grayscale histogram; Based on the local entropy of the bright / dark regions, an adaptive gradient entropy fusion weight is constructed. The mathematical model is as follows: ; in, This indicates that the weights are fused using adaptive gradient entropy. Represents the characteristic compensation coefficient. Represents a symbolic function. .
[0020] Step S42: Perform highlight-guided shadow enhancement and shadow-guided highlight enhancement to achieve mutual enhancement of details in both channels, obtaining the mutually enhanced features. The mathematical model is as follows: ; in, Indicates the features after mutual reinforcement. and These represent corrosion and expansion operations, respectively. Represents a structural element. , This indicates the closing operation.
[0021] Step S43: Highlight features Dark features and mutually reinforcing features Dynamic fusion is performed to obtain enhanced intelligent digital images for urology. The mathematical model is as follows: ; in, This represents an enhanced intelligent digital image of urology. The mathematical model for enhancing compensation weights is as follows: .
[0022] Preferably, the dual-domain feature dynamic fusion module constructed through S4 has the following advantages: Step S41 dynamically generates fusion weights based on the difference between the local gradient magnitude and information entropy of bright and dark features using local entropy and feature gradients, achieving intelligent balance for different brightness regions and improving structural coherence and detail preservation; Step S42, based on morphological operations, accurately extracts and strengthens weak tissue structures through mutual enhancement of bright and dark features, providing enhanced detail expression for subsequent feature fusion; Step S43 utilizes fusion weights to achieve precise fusion of bright features, dark features, and mutually enhanced features, ensuring the final fusion result maintains a natural transition while enhancing the visibility of pathological features, thus improving the clinical readability of intelligent digital images for urology. In summary, this module can significantly improve the robustness of detail enhancement, structural preservation ability, and synergistic performance of bright and dark features in intelligent digital images for urology.
[0023] Preferably, in S5, the construction of a urological intelligent digital image enhancement model specifically includes the following steps: Step S51: The original urological intelligent digital image is processed by the local statistical driven feature module to obtain multi-scale fusion features. The mathematical model is as follows: ; in, Indicates multi-scale fusion features, This represents a local statistically driven feature module. This represents the original intelligent digital image of urology.
[0024] Step S52: After the multi-scale fusion features are processed by the region weighted separation and denoising module, the bright and dark features based on mask reconstruction are obtained. The mathematical model is as follows: ; in, These represent the features of the highlights and the features of the shadows, respectively. This indicates the region-weighted separation and noise reduction module.
[0025] Step S53: The bright and dark features are processed by the dual-domain feature dynamic fusion module to obtain the enhanced intelligent digital image for urology. The mathematical model is as follows: ; in, This represents an enhanced intelligent digital image of urology. This indicates a dual-domain feature dynamic fusion module.
[0026] In summary, compared with existing technologies, the beneficial effects of this invention, due to the adoption of this technical solution, are as follows: The local statistical-driven feature module, through adaptive window multi-scale weighted fusion technology, can intelligently extract and fuse multi-scale features in medical images, improving the ability to express details and structures; the region weighted separation denoising module, through adaptive weighted total variation denoising technology, effectively removes noise and preserves the structural information of the image, ensuring high-quality image details; the dual-domain feature dynamic fusion module, through dual-domain feature dynamic fusion, optimizes the detail enhancement of bright and dark areas, improving the contrast between light and dark areas and the visualization of pathological features; these three modules together constitute a highly efficient intelligent digital image enhancement model for urology, enabling the model to intelligently balance denoising and detail enhancement in different brightness areas, significantly improving the clinical readability, structural preservation ability, and detail representation of intelligent digital images for urology. Attached Figure Description
[0027] Figure 1 This diagram illustrates the steps of an adaptive feature optimization and enhancement method for urinary system images.
[0028] Figure 2 This is a structural diagram of the local statistical-driven feature module.
[0029] Figure 3 This is a structural diagram of the region-weighted separation and denoising module.
[0030] Figure 4 This is a structural diagram of the dual-domain feature dynamic fusion module.
[0031] Figure 5 This is a diagram illustrating the overall structure of an intelligent digital image enhancement model for urology.
[0032] Figure 6 A schematic diagram illustrating the comparison of intelligent digital images in urology before and after surgery. Detailed Implementation
[0033] 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.
[0034] Please see the appendix Figure 1 -Appendix Figure 6 This invention provides a method for adaptive feature optimization and enhancement of urinary system images.
[0035] As attached Figure 1As shown in the flowchart, this invention proposes an adaptive feature optimization and enhancement method for urinary system images. The specific implementation method includes the following steps: S1. Acquire original X-ray images of the urinary system, preprocess the original images using a scanner, and construct a digital image dataset of the urinary system.
[0036] Furthermore, as shown in the appendix Figure 1 The intelligent digital image dataset for urology described in S1 specifically involves using ultrasound medical imaging equipment to capture images of selected areas, acquiring images according to clinical needs, and obtaining raw image data of three organs: kidney, bladder, and prostate. There are 560 kidney images, 400 bladder images, and 400 prostate images, totaling 1360 images. The analog images are converted into digital format using a scanner, and the acquired raw images are specifically cropped to a standard size of 320×320 pixels.
[0037] S2. Design an adaptive local window size and construct a local statistical driven feature module. The original urological intelligent digital image obtains multi-scale fusion features through the local statistical driven feature module.
[0038] Furthermore, as shown in the appendix Figure 1 The construction steps of the local statistical driven feature module described in S2 are as follows: Figure 2 As shown, the specific implementation of the module includes the following steps: Further, in step S21, design an adaptive local window size; the mathematical model is as follows: ; in, To adapt to the local window size, The parameters are learnable and updated through backpropagation. To normalize the activation function, compress the output to... , Original intelligent digital images for urology.
[0039] Further, in step S22, the values within the local window are statistically analyzed, and the local mean is calculated. The mathematical model is as follows: ; in It is a local mean. Represented as pixels within this local window, with a position of The mathematical model for calculating local variance is as follows: ; in, To calculate the local variance and local entropy, the mathematical model is as follows: ; in, For local entropy, This represents the normalized probability distribution of pixel values within a local window. Based on local mean, local variance, and local entropy, an adaptive multi-scale weighted coefficient is designed and generated. The mathematical model is as follows: ; in, and These are all adjustable parameters used to adjust the effects of local variance and local entropy, with initial values of [values to be filled in]. , ,in It is a Gaussian scale with a value of 3.
[0040] Further, in step S23, based on the adaptive multi-scale weighting coefficients, the multi-scale fusion features are output, and the mathematical model is as follows: ; in The standard deviation is The Gaussian kernel has a scale of 3. For convolution operations, Original intelligent digital images for urology, This is a multi-scale fusion feature.
[0041] S3. Design an adaptive noise weighting factor and construct a region weighted separation denoising module. Multi-scale fusion features are obtained by the region weighted separation denoising module to obtain bright and dark features based on mask reconstruction.
[0042] Furthermore, as shown in the appendix Figure 1 The construction steps of the region-weighted separation and denoising module described in S3 are as follows: Figure 3 As shown, the specific implementation of the module includes the following steps: Further, in step S31, the adaptive noise weighting factor is designed and implemented. The mathematical model is as follows: ; in, For window size Inner multi-scale fusion features standard deviation This is an adaptive noise weighting factor. This is a noise reduction control factor used to control the intensity of noise removal. The mathematical model for denoising and optimizing multi-scale fusion features is as follows: ; in, To optimize the output structural composition image, The magnitude of the image gradient. The amplitude is used; texture components are calculated based on the optimized output structural component image, and the mathematical model is as follows: ; in, These are the texture detail features obtained after difference.
[0043] Further, in step S32, the brightness threshold is calculated using the following mathematical model: ; in, The brightness threshold. The mean of the entire structural component image is given. The standard deviation of the entire structural component image is given. For adjustment coefficients, The mathematical model for calculating the region mask is as follows: ; ; in, For masking dark areas, This is a mask for the bright areas, allocated on a pixel-by-pixel basis.
[0044] Furthermore, in step S33, the bright and dark areas are reconstructed through weighted fusion, and the mathematical model is as follows: ; ; in, For the bright part features after reconstruction based on the mask, This represents the dark area features after reconstruction based on the mask.
[0045] S4. Construct adaptive gradient entropy fusion weights and dual-channel detail mutual enhancement, construct a dual-domain feature dynamic fusion module, and obtain the enhanced intelligent digital image of urology based on the bright and dark features after mask reconstruction through the dual-domain feature dynamic fusion module.
[0046] Further, in step S41, the bright part features are calculated. and dark features The characteristic gradient, mathematical model of which is: ; in, Represents the feature gradient. and These represent the gradients in the horizontal and vertical directions, respectively. Then calculate the highlights features. and dark features The local entropy, mathematical model is: ; in, This represents the local entropy of the bright / dark regions. Represents grayscale value The probability distribution of grayscale values, ; in Indicates the local window radius. This indicates the number of pixels within a local window. Represents a grayscale histogram; Based on the local entropy of the bright / dark regions, an adaptive gradient entropy fusion weight is constructed. The mathematical model is as follows: ; in, This indicates that the weights are fused using adaptive gradient entropy. Represents the characteristic compensation coefficient. , Represents a symbolic function. .
[0047] Further, in step S42, highlight-guided shadow enhancement and shadow-guided highlight enhancement are performed to achieve mutual enhancement of details in both channels, resulting in mutually enhanced features. The mathematical model is as follows: ; in, Indicates the features after mutual reinforcement. and These represent corrosion and expansion operations, respectively. Represents a structural element. , This indicates the closing operation.
[0048] Further, in step S43, the bright area features... Dark features and mutually reinforcing features Dynamic fusion is performed to obtain enhanced intelligent digital images for urology. The mathematical model is as follows: ; in, This represents an enhanced intelligent digital image of urology. The mathematical model for enhancing compensation weights is as follows: ; As attached Figure 1 The construction steps of the intelligent digital image enhancement model for urology described in S5 are attached. Figure 5As shown, the specific implementation of the model includes the following steps: Furthermore, in step S51, the original intelligent digital image of urology is processed by the local statistical driven feature module to obtain multi-scale fusion features. The mathematical model is as follows: ; in, Indicates multi-scale fusion features, This represents a local statistically driven feature module. This represents the original intelligent digital image of urology.
[0049] Furthermore, in step S52, the multi-scale fusion features are processed by a region-weighted separation and denoising module to obtain bright and dark features based on mask reconstruction. The mathematical model is as follows: ; in, These represent the features of the highlights and the features of the shadows, respectively. This indicates the region-weighted separation and noise reduction module.
[0050] Furthermore, in step S53, the bright and dark features are processed by the dual-domain feature dynamic fusion module to obtain the enhanced intelligent digital image for urology. The mathematical model is as follows: ; in, This represents an enhanced intelligent digital image of urology. This indicates a dual-domain feature dynamic fusion module.
[0051] A urological intelligent digital image enhancement model is trained by inputting the urological intelligent digital image to be enhanced into the trained urological intelligent digital image enhancement model and outputting the enhanced urological intelligent digital image.
[0052] Furthermore, as shown in the appendix Figure 1 The intelligent digital image enhancement model for urology described in S6, input with an intelligent digital image for urology, as shown in the attached image. Figure 6 As shown in (a), the enhanced intelligent digital image of urology is output, as attached. Figure 6 As shown in (b), the specific implementation includes the following steps: Furthermore, in step S6, the model uses the Debian operating system, Python 3.11.2, an Intel Xeon Platinum processor, OpenCV and PIL image processing libraries, and 128GB of physical memory on the server. The dataset contains 600 raw images of three organs: kidney, bladder, and prostate. During training, the mean squared error loss function is used as the optimization objective, with an initial learning rate of 0.001. As training progresses, the learning rate is decreased by a factor of 0.1 every 20 rounds. Backpropagation is performed using the Adam optimizer to progressively update the adaptive local window size. The parameters enable the model to gradually acquire the ability to accurately enhance the bright and dark areas and detailed features of urological images.
[0053] The above are merely preferred embodiments of the present invention. It should be noted that those skilled in the art can make various modifications and improvements without departing from the inventive concept of the present invention, and these modifications and improvements all fall within the protection scope of the present invention.
Claims
1. A method for adaptive feature optimization and enhancement of urinary system images, characterized in that, include: Acquire original X-ray images of the urinary system, preprocess the original images using a scanner, and construct a digital image dataset of the urinary system; Based on the adaptive local window size, an adaptive multi-scale weighting coefficient is designed, a local statistical driving feature module is constructed, and the digital image of the urinary system is processed according to the local statistical driving feature module to generate multi-scale fusion features of the image. The multi-scale fusion features are used by a region-weighted separation and denoising module to obtain the bright and dark features of the urinary system image after mask reconstruction; the region-weighted separation and denoising module is implemented based on an adaptive noise weighting factor. Based on the adaptive gradient entropy fusion weights, a dual-domain feature dynamic fusion module is constructed. The bright and dark features are used to generate an enhanced digital image of the urinary system through the dual-domain feature dynamic fusion module. A digital image enhancement model for the urinary system is constructed based on a local statistical driving feature module, a region weighted separation denoising module, and a dual-domain feature dynamic fusion module. The enhancement model is trained using the constructed dataset and outputs a high-resolution digital image of the urinary system.
2. The method for adaptive feature optimization and enhancement of urinary system images according to claim 1, characterized in that, For each pixel in the original urological digital image, the numerical difference between that pixel and its neighboring pixels is combined with learnable parameters to obtain the output. The output is then input into a normalized activation function to generate an adaptive local window size for the corresponding location. .
3. The method for adaptive feature optimization and enhancement of urinary system images according to claim 2, characterized in that, Within the adaptive local window, a local mean is constructed based on the combination relationship between the original image value at each location and its corresponding weight; Local variance is formed based on the local mean; the distribution of pixel values within the local window is normalized to a probability distribution, and a local entropy is constructed by combining the logarithmic term; two adjustable parameters are introduced to adjust the local variance and local entropy respectively, and the results at different scales are exponentially mapped by combining a set Gaussian scale, and an adaptive multi-scale weighted coefficient is constructed by normalization. ; Based on adaptive multi-scale weighting coefficients, features from all scales are weighted and combined to obtain the multi-scale fused features of the pixel. .
4. The method for adaptive feature optimization and enhancement of urinary system images according to claim 3, characterized in that, The standard deviation is extracted based on the distribution of multi-scale fusion features within an adaptive local window size. The standard deviation is then weighted and combined with a preset denoising control factor to obtain the response term. ; Perform an inverse proportional mapping operation on the response term to generate the final weighting factor used to characterize the noise level. ; Based on the difference between the original intelligent digital image of urology and the multi-scale fusion features, the adaptive noise weighting factor is introduced to establish a weighting, and the response results of all pixel locations are combined to form a structural component image. ; By combining the differences between the multi-scale fusion features and the structural component image, texture detail features are constructed. .
5. The method for adaptive feature optimization and enhancement of urinary system images according to claim 4, characterized in that, Based on the structural component image, the corresponding full-image mean and full-image standard deviation are extracted; an adjustment coefficient is introduced and combined with the full-image standard deviation, and a relationship is established with the full-image mean to construct a brightness threshold. ; Based on the comparison between the pixel values in the structural component image and the brightness threshold, a bright region mask is generated by dividing the image into pixels. Mask of dark areas .
6. The method for adaptive feature optimization and enhancement of urinary system images according to claim 5, characterized in that, Based on the combination of the bright region mask and the structural component image, and by combining the bright region mask with texture detail features, the bright region features after mask reconstruction are constructed. ; Based on the combination of dark region mask and structural component image, and by combining dark region mask with texture detail features, dark features after mask reconstruction are constructed. .
7. The method for adaptive feature optimization and enhancement of urinary system images according to claim 6, characterized in that, Based on the orientation of each pixel in the bright part features, the gradient information in the horizontal and vertical directions is extracted, and the gradient magnitude of the bright part features is constructed by combining the gradients in the two directions; the directional gradient information in the dark part features is extracted in the same way to construct the gradient magnitude of the dark part features. The distribution of gray values in the neighborhood of each pixel is statistically analyzed, gray value probabilities are constructed based on the frequency of occurrence, and entropy expressions are established by combining logarithmic terms to form the local entropy of bright and dark areas respectively. By incorporating the sign information between the feature compensation coefficient and the entropy difference between bright and dark regions, and combining this with the local entropy of the bright area, an adaptive gradient entropy fusion weight is constructed. .
8. The method for adaptive feature optimization and enhancement of urinary system images according to claim 7, characterized in that, Guided by the bright features, erosion and dilation sequence operations are performed on the dark features. The erosion and dilation are constructed based on preset structural units to guide the formation of bright-driven dark enhancement information. Guided by the dark features, erosion and dilation operations are sequentially performed on the bright features according to the same structural units to obtain dark-driven bright enhancement information. Based on the aforementioned dual-channel features, a mutual enhancement feature is obtained by combining highlight enhancement and shadow enhancement content. .
9. The method for adaptive feature optimization and enhancement of urinary system images according to claim 8, characterized in that, The adaptive gradient entropy fusion weights are mapped and combined with pi to guide the generation of a combined result; simultaneously, the mutually enhanced features are normalized to generate a combined result. The aforementioned two combined results are combined and normalized to form the enhancement compensation weights. ; The adaptive gradient entropy fusion weights are used to dynamically adjust the bright and dark features respectively, and the enhancement compensation weights are introduced to weight and mutually enhance the features. The three are then weighted and fused to generate an enhanced intelligent digital image for urology. .
10. The method for adaptive feature optimization and enhancement of urinary system images according to claim 9, characterized in that, The raw X-ray images of the urinary system are processed by a local statistical-driven feature module to obtain multi-scale fusion features; The multi-scale fusion features are processed by a region-weighted separation and denoising module to obtain bright and dark features based on mask reconstruction. Bright and dark features are dynamically fused through a dual-domain feature fusion module to obtain an enhanced intelligent digital image for urology.