A breast CT image enhancement method and system

Abstract

The present application relates to the field of medical images, in particular to a breast CT image enhancement method and system, specifically, based on the local structure tensor eigenvalue information and normalized local entropy information of the guide image of the original breast CT image, a composite edge perception weight is calculated; the original breast CT image is weighted and guided filtering by using the composite edge perception weight to obtain a detail layer image; the bright detail pixels in the detail layer image and the bright detail area formed by the bright detail pixels are identified, the morphological feature parameters of the bright detail area and the bright detail feature entropy in the local neighborhood of the bright detail pixels are calculated; based on the morphological feature parameters and the bright detail feature entropy, the bright detail pixels are selectively enhanced to obtain an optimized detail layer image; based on the image local statistical characteristics and the perception measurement based on the human visual system model, a fusion weight is calculated, and the base layer image and the optimized detail layer image are adaptively fused according to the fusion weight to obtain the final enhanced breast CT image.

Description

Technical Field

[0001] This invention relates to the field of medical imaging, specifically to a method and system for enhancing breast CT images. Background Technology

[0002] Breast cancer is one of the most common malignant tumors in women worldwide, and early detection and accurate diagnosis are crucial for improving patient survival rates. Compared to traditional two-dimensional mammography, breast CT provides non-overlapping tomographic images, which helps to display the three-dimensional structure, morphology, and edge features of lesions, especially advantageous in dense breasts. During the imaging process, breast CT images are limited by factors such as X-ray dose, detector performance, and reconstruction algorithms. The density differences between soft tissues (such as glands, fat, and fibrous tissue) are small, resulting in low overall image contrast. This is especially true for some early, small lesions or diffuse lesions, where the contrast with surrounding normal tissue may be very low, making them difficult to identify effectively. Furthermore, to reduce radiation dose, breast CT typically uses lower X-ray exposure, which inevitably leads to higher noise levels in the images. Especially in low-dose scanning modes, noise can severely affect image quality, obscuring minute structural and lesion details.

[0003] Conventional image enhancement techniques, such as histogram equalization, spatial domain filtering, transform domain methods, and partial differential equation methods, can improve edges and details, but they either amplify image noise or lose some grayscale information, both of which can affect medical personnel's judgment. Finding a way to suppress noise while preserving and sharpening edge details, especially microcalcifications, is of paramount importance for improving the early detection rate and diagnostic accuracy of breast cancer. Summary of the Invention

[0004] To address the aforementioned problems, in a first aspect, the present invention provides a method for enhancing breast CT images, the method comprising:

[0005] The original breast CT image is acquired, and a composite edge-aware weight is calculated based on the local structural tensor feature value information and normalized local entropy information of the guided image. The composite edge-aware weight is then used to perform weighted guided filtering on the original breast CT image to obtain a detail layer image.

[0006] Identify the bright detail pixels and the bright detail regions they form in the detail layer image, calculate the morphological feature parameters of the bright detail regions, and the bright detail feature entropy in the local neighborhood of the bright detail pixels; based on the morphological feature parameters and the bright detail feature entropy, selectively enhance the bright detail pixels to obtain an optimized detail layer image;

[0007] The fusion weights are calculated based on the local statistical characteristics of the image and the perceptual measurement based on the human visual system model. The base layer image and the optimized detail layer image are adaptively fused according to the fusion weights to obtain the final enhanced breast CT image.

[0008] Preferably, the composite edge-aware weight is calculated based on the local structural tensor feature value information and normalized local entropy information of the guiding image, specifically as follows:

[0009] Calculate the structure tensor of the guiding image in the neighborhood of each pixel, and solve for its eigenvalues;

[0010] Calculate the normalized local entropy of the guide image in the neighborhood of each pixel;

[0011] The composite edge-aware weights are calculated based on the eigenvalues ​​and the normalized local entropy.

[0012] Preferably, the selective enhancement of bright detail pixels based on morphological feature parameters and bright detail feature entropy specifically involves:

[0013] Pixels with brightness greater than the brightness threshold of bright detail pixels are obtained and connected component analysis is performed to obtain the bright detail region;

[0014] Calculate the morphological feature parameters of each bright detail region, the parameters including at least one of area, roundness, and density;

[0015] Calculate the local bright detail feature entropy of each bright detail pixel, and obtain the target range of morphological feature parameters and the target interval of entropy values; when the morphological feature parameters of the region to which the pixel belongs are within the target range and the bright detail feature entropy is within the target interval, apply a first enhancement rule to the pixel; otherwise, apply a second enhancement rule, wherein the enhancement intensity corresponding to the first enhancement rule is greater than that of the second enhancement rule.

[0016] Preferably, the fusion weights used in the adaptive fusion process include base layer weights and detail layer weights;

[0017] Calculate the local variance of the base layer image at pixel i and a perceptual metric based on a human visual system model, the metric including at least one of local contrast sensitivity or visual masking effect.

[0018] The detail layer weights are calculated using preset rules based on the local variance, the strength of the optimized detail layer values, and the perception metric.

[0019] Preferably, the method further includes suppressing non-bright detail pixels in the detail layer image.

[0020] In a second aspect of the invention, a breast CT image enhancement system is provided, the system comprising:

[0021] The guided filtering module is used to acquire the original breast CT image, calculate the composite edge-aware weight based on the local structural tensor feature value information and normalized local entropy information of the guided image, and use the composite edge-aware weight to perform weighted guided filtering on the original breast CT image to obtain the detail layer image.

[0022] The detail enhancement module is used to identify bright detail pixels and the bright detail regions formed in the detail layer image, calculate the morphological feature parameters of the bright detail regions, and the bright detail feature entropy in the local neighborhood of the bright detail pixels; and selectively enhance the bright detail pixels based on the morphological feature parameters and the bright detail feature entropy to obtain an optimized detail layer image.

[0023] The fusion module is used to calculate the fusion weight based on the local statistical characteristics of the image and the perceptual metric based on the human visual system model. Based on the fusion weight, the base layer image and the optimized detail layer image are adaptively fused to obtain the final enhanced breast CT image.

[0024] Preferably, the composite edge-aware weight is calculated based on the local structural tensor feature value information and normalized local entropy information of the guiding image, specifically as follows:

[0025] Calculate the structure tensor of the guiding image in the neighborhood of each pixel, and solve for its eigenvalues;

[0026] Calculate the normalized local entropy of the guide image in the neighborhood of each pixel;

[0027] The composite edge-aware weights are calculated based on the eigenvalues ​​and the normalized local entropy.

[0028] Preferably, the selective enhancement of bright detail pixels based on morphological feature parameters and bright detail feature entropy specifically involves:

[0029] Pixels with brightness greater than the brightness threshold of bright detail pixels are obtained and connected component analysis is performed to obtain the bright detail region;

[0030] Calculate the morphological feature parameters of each bright detail region, the parameters including at least one of area, roundness, and density;

[0031] Calculate the local bright detail feature entropy of each bright detail pixel, and obtain the target range of morphological feature parameters and the target interval of entropy values; when the morphological feature parameters of the region to which the pixel belongs are within the target range and the bright detail feature entropy is within the target interval, apply a first enhancement rule to the pixel; otherwise, apply a second enhancement rule, wherein the enhancement intensity corresponding to the first enhancement rule is greater than that of the second enhancement rule.

[0032] Preferably, the fusion weights used in the adaptive fusion process include base layer weights and detail layer weights;

[0033] Calculate the local variance of the base layer image at pixel i and a perceptual metric based on a human visual system model, the metric including at least one of local contrast sensitivity or visual masking effect.

[0034] The detail layer weights are calculated using preset rules based on the local variance, the strength of the optimized detail layer values, and the perception metric.

[0035] Finally, the present invention provides a computer-readable storage medium storing a computer program, characterized in that the computer program, when executed by a processor, implements the method described in the first aspect.

[0036] This invention employs a weighted guided filtering method that combines structural tensor eigenvalues ​​and local entropy for edge perception. This method more accurately distinguishes between real edges, textures, and noise, thereby effectively smoothing noise while more precisely preserving or even sharpening tissue boundaries and lesion contours in the image. Furthermore, by combining morphological analysis and entropy analysis of bright details, it achieves highly selective enhancement of bright details with specific morphological and low-entropy characteristics, while suppressing noise and irrelevant textures, thus improving the signal-to-noise ratio of key diagnostic information. Attached Figure Description

[0037] Figure 1 This is a flowchart of Example 1;

[0038] Figure 2 This is a diagram illustrating the original image, the base layer, and the detail layer.

[0039] Figure 3 Before and after comparison images for optimizing detail layers;

[0040] Figure 4 Before and after optimization of the original image;

[0041] Figure 5 This is a structural diagram of Example 2. Detailed Implementation

[0042] In this document, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, without necessarily requiring or implying any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0043] 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.

[0044] Example 1, as Figure 1 The breast CT image enhancement methods shown include:

[0045] S1. Obtain the original breast CT image, calculate the composite edge-aware weight based on the local structural tensor feature value information and normalized local entropy information of the guided image; use the composite edge-aware weight to perform weighted guided filtering on the original breast CT image to obtain the detail layer image;

[0046] Acquire the raw breast CT image I to be processed, and set the guide image G, preferably G = I. Calculate the composite edge-aware weight F. * In one embodiment, the composite edge-aware weight is calculated based on the local structural tensor feature value information and normalized local entropy information of the guiding image, specifically as follows:

[0047] Calculate the structure tensor of the guiding image in the neighborhood of each pixel, and solve for its eigenvalues;

[0048] Calculate the normalized local entropy of the guide image in the neighborhood of each pixel;

[0049] The composite edge-aware weights are calculated based on the eigenvalues ​​and the normalized local entropy.

[0050] For the guiding image G, compute its structure tensor j in the neighborhood of each pixel i. iThe structure tensor is a matrix constructed based on image gradient information within the pixel neighborhood, reflecting the structural pattern of the local image. Solving for the eigenvalues ​​λ of this structure tensor... j,1 and λ j,2 , λ j,1 ≥λ j,2 If both eigenvalues ​​are small, it indicates a flat region; if λ... j,1 Much greater than λ j,2 , indicates strong edges or linear structures; both larger indicate corner points or complex textures.

[0051] Furthermore, the normalized local entropy H of the guiding image G in the neighborhood of each pixel i is calculated. norm,i Local entropy represents the randomness or complexity of grayscale value distribution within a pixel's neighborhood. High entropy values ​​correspond to regions with rich texture or high noise, while low entropy values ​​correspond to regions with uniform or flat grayscale values. Based on the feature value λ... j,1 and λ j,2 and normalized local entropy H norm,i Calculate the composite edge-aware weights In one embodiment, when the eigenvalue indicates the presence of a strong and simple edge structure, such as λ j,1 Large and much larger than λ j,2 And the local entropy H norm,i When the eigenvalue is low, a higher weight value is assigned; when the eigenvalue indicates a flat region, such as λ... j,1 , λ j,2 Small values ​​or a combination of eigenvalues ​​and entropy values ​​indicate a noisy / complex texture region, such as H. norm,i When the value is high, a lower weight value is assigned. This invention combines a weight calculation method based on geometric structure and information complexity. Compared with traditional methods that rely solely on variance, it can more robustly and accurately reflect the edge preservation requirements at pixel locations. It overcomes the shortcomings of traditional weights, which are based solely on the degree of brightness change and are susceptible to noise interference, and have difficulty distinguishing between texture and edges.

[0052] For example, if a pixel is located on a clear tissue boundary, the calculated feature value is λ. j,1 =100,λ j,2 =5, entropy H norm,i =0.1, these values ​​indicate strong edges, simple structure, and calculated composite weights It will be very high, for example, close to 1. If the pixel is located inside uniform adipose tissue, then the feature value is λ. j,1 =2,λ j,2 =1, entropy H norm,i =0.05 indicates a flat region, and the calculated weights are... The value will be lower. If the pixel is located in a noisy region, the feature value may be uncertain, but the entropy H will be lower. norm,i A value of 0.8 would be very high; high entropy would lead to errors in the calculated weights. The value is very low, so that different processing intensities can be applied to different regions during subsequent filtering.

[0053] Using the calculated composite edge-sensing weight Γ * A weighted guided filtering operation is performed on the original breast CT image I. It should be noted that in this invention, the direct output of the weighted guided filtering is a smoothed version of the image, i.e., the base layer image B. This base layer B uses Γ... * Weighting can effectively smooth out noise while maintaining the properties of Γ. * The important edge structures were identified. Then, the detail layer image D = IB was obtained by subtracting the original image I from the base layer B, as shown below. Figure 2 As shown. The detail layer D contains high-frequency information separated from the original image, including noise and various fine structures.

[0054] S2, identify the bright detail pixels and the bright detail regions formed in the detail layer image, calculate the morphological feature parameters of the bright detail regions and the bright detail feature entropy in the local neighborhood of the bright detail pixels; selectively enhance the bright detail pixels based on the morphological feature parameters and the bright detail feature entropy to obtain the optimized detail layer image;

[0055] Identify all brightness values ​​in D that are greater than the brightness threshold T. bright Pixels that are not clearly defined are labeled as bright detail pixels. Connectivity analysis is performed on these bright detail pixels to identify regions composed of interconnected bright pixels, i.e., bright detail regions. For each identified bright detail region, its morphological feature parameters are calculated, such as the region's area, roundness, and density. The density is the ratio of the average brightness of pixels within the region to the maximum brightness, or the ratio of the area to the area of ​​the circumscribed rectangle.

[0056] For each bright detail pixel i, calculate the local bright detail feature entropy H of the distribution of brightness values ​​of other bright detail pixels in its local neighborhood (e.g., a 3x3 or 5x5 window). BDFE,i Low entropy indicates high brightness consistency among bright pixels in the neighborhood. Then, selective enhancement is performed by pre-setting target ranges for morphological parameters (e.g., typical area range corresponding to microcalcifications, lower limit of roundness, lower limit of density) and target intervals for entropy values ​​(e.g., low entropy intervals). For each bright detail pixel i: determine whether the morphological parameters of its region fall within the target range, and whether its own H... BDFE,i Does it fall within the low-entropy range? If both conditions are met—that is, the morphology resembles micro-calcifications and the internal structure is homogeneous—then the pixel is considered highly likely to be a feature of interest, and the first enhancement rule is applied, i.e., strong enhancement is performed, for example, multiplying its brightness value by a large enhancement coefficient k. strong>1. If at least one condition is not met—that is, irregular shape, inconsistent size, or large internal brightness variation—it is more likely to be considered noise or ordinary texture. A second enhancement method is then applied, either weak enhancement or no enhancement, for example, multiplying by a small coefficient k. weak Preferably, 1≤k weak <k strong , or k weak =1. In an optional embodiment, for non-bright detail pixels in detail layer D, i.e., those with a brightness value lower than T... bright If the brightness value is too low, suppression processing is applied, such as multiplying it by a suppression factor less than 1, or thresholding to remove weak noise. After completing the above operations, the optimized detail layer image D′ is obtained, as shown below. Figure 3 As shown.

[0057] S3. Based on the local statistical characteristics of the image and the perceptual measurement based on the human visual system model, the fusion weight is calculated. The base layer image and the optimized detail layer image are adaptively fused according to the fusion weight to obtain the final enhanced breast CT image.

[0058] In another embodiment, the local statistical characteristics of the image required for fusion are calculated. These local statistical characteristics include the local variance of the base layer B at each pixel i, and / or the intensity information of the optimized detail layer D′ at pixel i, such as its absolute value |D′| or its value after function mapping. Then, a perceptual metric based on the HVS model is calculated. This perceptual metric includes, but is not limited to, local contrast sensitivity and visual masking effect metrics. In yet another embodiment, the contrast sensitivity is estimated based on the spatial frequency characteristics of the neighborhood of pixel i to determine the human eye's sensitivity to contrast changes at that location. The visual masking effect metric is estimated based on the texture complexity or edge intensity of the neighborhood of pixel i to determine the degree to which a strong background masks detail information.

[0059] The detail layer fusion weight β is calculated based on the local statistical characteristics and HVS perceptual metric. i When the local variance of the base layer is large and the strength of the detailed layer is high after optimization, increase β. i If the human eye has high contrast sensitivity and a weak masking effect at that location, then β can be maintained or further enhanced. i Conversely, if the sensitivity is low or the masking effect is strong, β will be significantly reduced. i This is because, in the latter case, injecting details may not help improve perception and could instead introduce visual interference. Calculate β. i Then, the weight of the base layer is α. i =1-β i According to I enhanced (i)=α i B i +βi D′ i The final enhanced image I is obtained. enhanced (i), such as Figure 4 As shown.

[0060] Example 2, as Figure 5 As shown, a breast CT image enhancement system is provided, the system comprising:

[0061] The guided filtering module is used to acquire the original breast CT image, calculate the composite edge-aware weight based on the local structural tensor feature value information and normalized local entropy information of the guided image, and use the composite edge-aware weight to perform weighted guided filtering on the original breast CT image to obtain the detail layer image.

[0062] The detail enhancement module is used to identify bright detail pixels and the bright detail regions formed in the detail layer image, calculate the morphological feature parameters of the bright detail regions, and the bright detail feature entropy in the local neighborhood of the bright detail pixels; and selectively enhance the bright detail pixels based on the morphological feature parameters and the bright detail feature entropy to obtain an optimized detail layer image.

[0063] The fusion module is used to calculate the fusion weight based on the local statistical characteristics of the image and the perceptual metric based on the human visual system model. Based on the fusion weight, the base layer image and the optimized detail layer image are adaptively fused to obtain the final enhanced breast CT image.

[0064] Preferably, the composite edge-aware weight is calculated based on the local structural tensor feature value information and normalized local entropy information of the guiding image, specifically as follows:

[0065] Calculate the structure tensor of the guiding image in the neighborhood of each pixel, and solve for its eigenvalues;

[0066] Calculate the normalized local entropy of the guide image in the neighborhood of each pixel;

[0067] The composite edge-aware weights are calculated based on the eigenvalues ​​and the normalized local entropy.

[0068] Preferably, the selective enhancement of bright detail pixels based on morphological feature parameters and bright detail feature entropy specifically involves:

[0069] Pixels with brightness greater than the brightness threshold of bright detail pixels are obtained and connected component analysis is performed to obtain the bright detail region;

[0070] Calculate the morphological feature parameters of each bright detail region, the parameters including at least one of area, roundness, and density;

[0071] Calculate the local bright detail feature entropy of each bright detail pixel, and obtain the target range of morphological feature parameters and the target interval of entropy values; when the morphological feature parameters of the region to which the pixel belongs are within the target range and the bright detail feature entropy is within the target interval, apply a first enhancement rule to the pixel; otherwise, apply a second enhancement rule, wherein the enhancement intensity corresponding to the first enhancement rule is greater than that of the second enhancement rule.

[0072] Preferably, the fusion weights used in the adaptive fusion process include base layer weights and detail layer weights;

[0073] Calculate the local variance of the base layer image at pixel i and a perceptual metric based on a human visual system model, the metric including at least one of local contrast sensitivity or visual masking effect.

[0074] The detail layer weights are calculated using preset rules based on the local variance, the strength of the optimized detail layer values, and the perception metric.

[0075] Embodiment 3 provides a computer-readable storage medium storing a computer program, characterized in that the computer program, when executed by a processor, implements the method described in Embodiment 1.

[0076] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of a necessary general-purpose hardware platform, or by a combination of hardware and software. Based on this understanding, the above technical solutions, in essence or in terms of their contribution to the prior art, can be embodied in the form of a computer product. This invention can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0077] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Other embodiments may also be used. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for enhancing breast CT images, characterized in that, The method includes: A raw breast CT image is acquired, and a composite edge-aware weight is calculated based on the local structural tensor eigenvalues ​​and normalized local entropy information of the guided image. The composite edge-aware weight is then used to perform weighted guided filtering on the raw breast CT image to obtain a detail layer image. The guided image is the raw CT breast image. Identify the bright detail pixels and the bright detail regions they form in the detail layer image, calculate the morphological feature parameters of the bright detail regions, and the bright detail feature entropy in the local neighborhood of the bright detail pixels; based on the morphological feature parameters and the bright detail feature entropy, selectively enhance the bright detail pixels to obtain an optimized detail layer image; The fusion weights are calculated based on the local statistical characteristics of the image and the perceptual measurement based on the human visual system model. The base layer image and the optimized detail layer image are adaptively fused according to the fusion weights to obtain the final enhanced breast CT image. The base layer image is a smoothed version of the image directly output by weighted guided filtering. The detail layer image is obtained by the difference operation between the original image and the base layer image. The composite edge-aware weight is calculated based on the local structure tensor feature value information and normalized local entropy information of the guiding image. Specifically, the structure tensor of the guiding image in each pixel neighborhood is calculated, and its feature value is solved; the normalized local entropy of the guiding image in each pixel neighborhood is calculated; and the composite edge-aware weight is calculated based on the feature value and the normalized local entropy.

2. The method as described in claim 1, characterized in that, Selective enhancement of bright detail pixels is performed based on morphological feature parameters and bright detail feature entropy, specifically as follows: Pixels with brightness greater than the brightness threshold of bright detail pixels are obtained and connected component analysis is performed to obtain the bright detail region; Calculate the morphological feature parameters of each bright detail region, the parameters including at least one of area, roundness, and density; Calculate the local bright detail feature entropy of each bright detail pixel, and obtain the target range of morphological feature parameters and the target interval of entropy values; when the morphological feature parameters of the region to which the pixel belongs are within the target range and the bright detail feature entropy is within the target interval, apply a first enhancement rule to the pixel; otherwise, apply a second enhancement rule, wherein the enhancement intensity corresponding to the first enhancement rule is greater than that of the second enhancement rule.

3. The method as described in claim 1, characterized in that, The fusion weights used in the adaptive fusion process include base layer weights and detail layer weights; Calculate the local variance of the base layer image at pixel i and a perceptual metric based on a human visual system model, the metric including at least one of local contrast sensitivity or visual masking effect. The detail layer weights are calculated using preset rules based on the local variance, the strength of the optimized detail layer values, and the perception metric.

4. The method as described in claim 1, characterized in that, The method also includes suppressing non-bright detail pixels in the detail layer image.

5. A breast CT image enhancement system, characterized in that, The system includes: A guided filtering module is used to acquire the original breast CT image, calculate a composite edge-aware weight based on the local structural tensor feature value information and normalized local entropy information of the guided image, and use the composite edge-aware weight to perform weighted guided filtering on the original breast CT image to obtain a detail layer image; the guided image is the original breast CT image. The detail enhancement module is used to identify bright detail pixels and the bright detail regions formed in the detail layer image, calculate the morphological feature parameters of the bright detail regions, and the bright detail feature entropy in the local neighborhood of the bright detail pixels; and selectively enhance the bright detail pixels based on the morphological feature parameters and the bright detail feature entropy to obtain an optimized detail layer image. The fusion module is used to calculate the fusion weight based on the local statistical characteristics of the image and the perceptual measurement based on the human visual system model. Based on the fusion weight, the base layer image and the optimized detail layer image are adaptively fused to obtain the final enhanced breast CT image. The composite edge-aware weight is calculated based on the local structure tensor feature value information and normalized local entropy information of the guiding image. Specifically, the structure tensor of the guiding image in each pixel neighborhood is calculated and its feature value is solved; the normalized local entropy of the guiding image in each pixel neighborhood is calculated; and the composite edge-aware weight is calculated based on the feature value and the normalized local entropy. The base layer image is a smoothed version of the image directly output by weighted guided filtering; the detail layer image is obtained by subtracting the original image from the base layer image.

6. The system as described in claim 5, characterized in that, Selective enhancement of bright detail pixels is performed based on morphological feature parameters and bright detail feature entropy, specifically as follows: Pixels with brightness greater than the brightness threshold of bright detail pixels are obtained and connected component analysis is performed to obtain the bright detail region; Calculate the morphological feature parameters of each bright detail region, the parameters including at least one of area, roundness, and density; Calculate the local bright detail feature entropy of each bright detail pixel, and obtain the target range of morphological feature parameters and the target interval of entropy values; when the morphological feature parameters of the region to which the pixel belongs are within the target range and the bright detail feature entropy is within the target interval, apply a first enhancement rule to the pixel; otherwise, apply a second enhancement rule, wherein the enhancement intensity corresponding to the first enhancement rule is greater than that of the second enhancement rule.

7. The system as described in claim 5, characterized in that, The fusion weights used in the adaptive fusion process include base layer weights and detail layer weights; Calculate the local variance of the base layer image at pixel i and a perceptual metric based on a human visual system model, the metric including at least one of local contrast sensitivity or visual masking effect. The detail layer weights are calculated using preset rules based on the local variance, the strength of the optimized detail layer values, and the perception metric.

8. A computer-readable storage medium storing a computer program thereon, characterized in that, When the computer program is executed by a processor, it implements the method as described in any one of claims 1-4.

Images