A glare suppression low-illumination image enhancement method and system based on global mapping and local tuning

Abstract

The application further provides a glare suppression low-illumination image enhancement method based on global mapping and local optimization, comprising the following steps: step S1: performing preliminary enhancement on an original image, and adjusting color saturation and contrast of the image; step S2: performing image region gradient detection on the original image to calculate regions with relatively sharp light and dark changes, and performing further enhancement on the image through a corresponding algorithm; and step S3: through a linear expansion algorithm, superimposing the images before and after enhancement through a pixel-level weight adaptive algorithm, thereby suppressing glare in the image, and adjusting contrast and color in the image after glare suppression through an algorithm, so that the image conforms to human eye visual habits. The technical scheme has the advantages of rapidness, good robustness, and adaptability to various dark and glare scenes, and can ensure good enhancement effect and real-time performance under various low-illumination scenes and image resolutions.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of image enhancement, in particular to a glare suppression low-light image enhancement method and system based on global mapping and local optimization. BACKGROUND

[0002] With the rapid development of the information age, it is essential to obtain information effectively and quickly. As an important information medium, images play an important role in people's daily life. The quality of images is required to be high in computer high-level vision applications, such as security monitoring, face recognition, target detection and intelligent driving, etc. However, due to the inevitable environmental factors, such as night, backlight, etc. Unideal lighting conditions, or limited by equipment defects and shooting technology, the obtained images have many problems, such as buried scene content, reduced contrast, serious noise pollution and inaccurate color, etc. These characteristics reduce the usability of such images, which not only seriously affects human subjective experience, but also greatly limits subsequent computer vision tasks. Therefore, it is of great significance to carry out research on low-light image enhancement methods and further obtain and utilize information in images.

[0003] Low-light images are often low in overall brightness, and the overall outline and detail texture of the image are difficult to identify, resulting in less information and lower contrast in the image. At the same time, noise can easily cover the detail texture of the image, so it is difficult for us to obtain good quality image data directly from low-light images.

[0004] Existing low-light image enhancement methods can be roughly divided into two categories: traditional methods and deep learning-based methods. The traditional method is usually based on simple and fixed rules and prior knowledge, and its enhancement effect is often unsatisfactory, which is easy to lose image information and details. And some methods usually need to be iterated several times to get a better enhancement result, so the real-time performance is poor. For deep learning methods, it is difficult to balance the enhancement effect and model size, so it is difficult to apply in practice. In addition, existing methods usually do not pay attention to the glare and spot problems in low-light images, and many methods usually amplify the spots and halos in the picture, causing poor visual experience. SUMMARY

[0005] Therefore, the purpose of the present application is to provide a glare suppression low-light image enhancement method and system based on global mapping and local optimization, which has the advantages of fast, good robustness, and adaptation to various dark and glare scenes. It can guarantee good enhancement effect and real-time performance in various low-light scenes and image resolutions.

[0006] To achieve the above object, the present application adopts the following technical scheme: a glare suppression low-illumination image enhancement method based on global mapping and local optimization, comprising the following steps:

[0007] Step S1: performing preliminary enhancement on the original image, performing global tone mapping on the original image by an improved algorithm parameter optimization strategy to restore certain details and information of the original image, and adjusting the color saturation and contrast of the image;

[0008] Step S2: performing image region gradient detection on the original image to calculate the region with relatively sharp light and dark changes, and performing further enhancement on the image by a corresponding algorithm in an adaptive manner;

[0009] Step S3: performing linear expansion algorithm, superimposing the images before and after enhancement by a pixel-level weight adaptive algorithm, thereby suppressing the glare in the image, and adjusting the contrast and color of the image after glare suppression in an adaptive manner by an algorithm, so that the image conforms to the habit of human eye vision.

[0010] In a preferred embodiment, the step S1 specifically comprises the following steps:

[0011] Step S11: converting the original image from the RGB space to the HSV space, and the specific calculation formula is as follows:

[0012] V = max (R, G, B) (1)

[0013]

[0014] Wherein, H represents the hue component of the HSV space, indicating the color type or color phase, S represents the saturation component of the HSV space, indicating the color brightness, and V represents the brightness component of the HSV space, indicating the color brightness;

[0015] Step S12: calculating the logarithmic average value Y of the brightness by using the V component calculated in step S11 ave :

[0016]

[0017] Wherein, MN represents the product of the height and width of the original image, σ represents a constant, x represents the horizontal coordinate of the pixel point, and y represents the vertical coordinate of the pixel point.

[0018] Step S13: calculating the standard deviation δ of all pixels in the V component by using formula (5), wherein v(i,j) represents a pixel on the V component, and μ represents the pixel mean value of the original image.

[0019]

[0020] Step S14: According to the different pixel standard deviation, the input original image is classified according to formula (6), and the input original image is classified into three categories: non-uniformly illuminated image, uniformly illuminated image and image with light-dark difference between the two;

[0021]

[0022] Wherein δ represents the input original image pixel standard deviation, δ1=0.17, δ2=0.09;

[0023] Step S15: According to the value of the pixel standard deviation δ, different L ave Adjustment is made by using different adjustment functions to enhance the curvature of the global tone mapping function curve, so as to enhance the adaptability to different types of dark scenes with different brightness conditions.

[0024] In a preferred embodiment, the step S1 further comprises:

[0025] Step S16: The optimization adjustment strategy of the luminance logarithmic mean can be shown by the following piecewise function, wherein Y c represents the luminance logarithmic mean after adjustment by the adjustment function;

[0026]

[0027] Step S17: Parameter P1 is introduced to adjust Y c and V max , and the adjustment formula is as follows:

[0028] Y cln =ln(P1Y c +1)

[0029] V maxln =ln(P1V max +1) (8)

[0030] Wherein Y cln represents the luminance logarithmic mean obtained after the above adjustment, V maxln represents the maximum luminance obtained after the above adjustment, and P1 represents an empirical parameter for adjusting the representation ability of the enhancement curve to improve the final effect of the present application, P1 is set to 13 in the present application;

[0031] Step S18: Y cln and V maxln obtained are brought into the following formula for calculation, and adjustment parameter P2 is introduced therein, wherein L e represents the result obtained after preliminary enhancement after global tone mapping;

[0032]

[0033] Step S19: using Newton iteration algorithm to calculate the threshold value V approaching to the enhanced curve intercept threshold value V due to P1 and P2 parameters cut , wherein the cycle parameter of Newton iteration method is set to 20 in the present application, and only the pixel part less than the threshold value can be enhanced by the enhanced curve, thus obtaining:

[0034]

[0035] In a preferred embodiment, the step S2 further comprises:

[0036] Step S21: using Scharr operator to perform convolution operation on the luminance component V obtained in the global tone mapping module, so as to extract the gradient information at each pixel point and obtain the gradient weight information maps in x direction and y direction;

[0037] Step S22: the size of Scharr operator is 3x3, and the difference between pixels is amplified by the weight coefficient in the filter, and the Scharr in x direction and y direction is defined as follows:

[0038]

[0039] Step S23: using the two Scharr operators in step S22 to perform convolution operation on the luminance component V, and obtaining two matrices g x and g y with the same size as the original image, and the specific calculation formula is as follows, wherein G x () represents the convolution operation on the image, and the convolution kernel is G x , G y () represents the convolution operation on the image, and the convolution kernel is G y ;

[0040] g x (x,y)=G x (V(x,y)),g y (x,y)=G y (V(x,y)) (12);

[0041] Step S24: obtaining the enhanced weight map of gradient information according to the following formula, wherein σ represents a very small constant to avoid the case that the gradient is 0;

[0042]

[0043] Step S25: The final enhanced weight map is obtained by combining the luminance component V with the re weight map obtained in the previous step and performing a power law transformation, where β = 255, and a is an empirical parameter;

[0044]

[0045] Step S26: A gradient detection local enhancement module is used to further enhance the image after global tone mapping, which can restore the image detail information. The image after local enhancement is obtained by the following formula:

[0046]

[0047] In a preferred embodiment, the step S3 specifically comprises:

[0048] Step S31: A linear expansion algorithm is used for glare suppression;

[0049] Step S32: A linear expansion algorithm that matches the human eye is used, which is specifically shown in the following formula, where L e represents the image after enhancement by the above two modules, L o represents the original image, and a1 and a2 represent the corresponding weights;

[0050] L p = a1L o (x, y) + a2L e (x, y) (16)

[0051] L p represents the result obtained after the linear expansion algorithm;

[0052] Step S33: The weights a1 and a2 in the linear expansion algorithm have different values at different brightness positions. In the brighter pixel area, the weight of the original image is higher, and in the darker pixel area, the weight of the original image is lower; where the threshold T = 256 - V ave , a2 = 1 - a1, where V ave represents the average value of the image brightness;

[0053]

[0054] Step S34: CLAHE algorithm is used to further improve the image contrast on the basis of preserving the glare suppression effect;

[0055] Step S35: The image is converted from RGB space to XYZ space, and then from XYZ space to Lab space, and then the CLAHE algorithm is applied to the L component. The formulas used for space conversion are shown in (18), (19), and (20):

[0056]

[0057]

[0058] In a preferred embodiment, step S3 further includes:

[0059] Step S36: Apply the CLAHE algorithm to the L component obtained in step S35, setting the contrast l in the CLAHE algorithm to 1 and the size of the divided sub-images to 10×10. After enhancement by the CLAHE algorithm, the final output image is obtained, as shown below, where L... Le This represents the L component after Le is transformed to Lab space;

[0060] L out =CLAHE(L Le (21)

[0061] Step S37: Combine the L component obtained after contrast processing of the image with the other unprocessed components and convert them back to the RGB space to obtain the final enhanced image of the present invention.

[0062] This invention also provides a glare suppression low-light image enhancement system based on global mapping and local optimization, which runs the aforementioned glare suppression low-light image enhancement method based on global mapping and local optimization; it also includes a global tone mapping enhancement module, a gradient detection local enhancement module, and a glare suppression module.

[0063] In a preferred embodiment, the global tone mapping enhancement module is responsible for initially enhancing the original image, performing global tone mapping through an improved algorithm parameter optimization strategy to restore certain details and information from the original image, and adjusting the color saturation and contrast of the image.

[0064] In a preferred embodiment, the gradient detection local enhancement module performs image region gradient detection on the original image to calculate the regions with more dramatic changes in brightness, and then adaptively enhances the image using a corresponding algorithm.

[0065] In a preferred embodiment, the glare suppression module uses a linear extension algorithm and a pixel-level weighted adaptive algorithm to superimpose the images before and after enhancement, thereby suppressing glare in the image. Furthermore, the algorithm adaptively adjusts the contrast and color in the image after glare suppression to make the image conform to human visual habits.

[0066] Compared with existing technologies, this invention has the following advantages: Compared with existing low-light image enhancement methods, this invention combines low-light image enhancement with glare suppression, taking into account various nighttime scenes and lighting conditions. It can not only effectively recover details in darker areas of the image, but also avoid amplifying existing bright spots and halos, and can suppress brighter areas in the image to a certain extent, achieving overall image brightness balance. Furthermore, this invention, while having good enhancement effects, also features high real-time performance, achieving a balance between enhancement effect and efficiency, and ensuring good real-time performance at various image resolutions. In summary, this invention can simultaneously handle the restoration tasks of various lighting scenarios, has good enhancement effects, and can guarantee good real-time performance for practical applications. Attached Figure Description

[0067] Figure 1 This is a schematic diagram of the technical route of a preferred embodiment of the present invention;

[0068] Figure 2 This is a flowchart illustrating a preferred embodiment of the present invention;

[0069] Figure 3 This is a graph showing the relationship between the logarithmic mean of brightness and the enhancement value in a preferred embodiment of the present invention.

[0070] Figure 4 This is a schematic diagram comparing the effects of the present invention (a preferred embodiment of the present invention) with other low-light enhancement methods. Detailed Implementation

[0071] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0072] It should be noted that the following detailed descriptions are illustrative and intended to provide further explanation of this application. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains.

[0073] It should be noted that the terminology used herein is for the purpose of describing particular implementations only and is not intended to limit the exemplary implementations according to this application; as used herein, the singular form is intended to include the plural form as well, unless the context clearly indicates otherwise; furthermore, it should be understood that when the terms “comprising” and / or “including” are used in this specification, they indicate the presence of features, steps, operations, devices, components and / or combinations thereof.

[0074] like Figures 1-4 As shown, the method of the present invention can be specifically divided into the following modules: global tone mapping enhancement module, gradient detection local enhancement module, and glare suppression module.

[0075] Global tone mapping module: This module is responsible for the initial enhancement of the original image. It uses an improved algorithm parameter optimization strategy to perform global tone mapping to restore certain details and information from the original image, and adjusts the color saturation and contrast of the image.

[0076] Gradient weight local enhancement module: This module performs image region gradient detection on the original image to calculate the regions with drastic changes in brightness, and then adaptively enhances the image using corresponding algorithms.

[0077] Glare suppression module: The task of this module is to suppress glare in the image by superimposing the images before and after enhancement through a linear expansion algorithm and a pixel-level weight adaptive algorithm. It also uses an algorithm to adaptively adjust the contrast and color in the image after glare suppression so that the image conforms to the visual habits of the human eye.

[0078] Specifically, the steps for the global tone mapping module are as follows:

[0079] 1) The three channels of the HSV color space represent the hue, saturation, and brightness of an image, respectively. Compared to the RGB color space, the HSV color space is more consistent with human color perception, and converting an image to the HSV color space allows for quick acquisition of the image's brightness component, which is beneficial for subsequent calculations in the module. The specific calculation formula for converting the original image from RGB color space to HSV color space is shown below:

[0080] V = max(R, G, B) (1)

[0081]

[0082] Wherein, H represents the hue component of the HSV space, referring to the type or hue of a color; S represents the saturation component of the HSV space, referring to the vividness of a color; and V represents the lightness component of the HSV space, referring to the brightness of a color.

[0083] 2) Calculate the logarithmic mean Y of the lightness using the V component obtained in the previous step. ave The calculation formula is shown below, where MN represents the product of the height and width of the original image, and σ represents a very small constant to avoid the original image having all pixels as black pixels with a pixel value of 0.

[0084]

[0085] Where MN represents the product of the height and width of the original image, σ represents a constant, x represents the horizontal coordinate of the pixel, and y represents the vertical coordinate of the pixel;

[0086] 3) The standard deviation of all pixels in a brightness image can reflect the uniformity of the image illumination to a certain extent. The standard deviation δ of all pixels in the V component is calculated using the following formula, where v(i,j) represents a pixel in the V component and μ represents the pixel mean of the original image.

[0087]

[0088] 4) Based on different pixel standard deviations, this invention classifies the input raw images according to the following formula, and categorizes the input raw images into three types: images with non-uniform illumination, images with uniform illumination, and images with brightness differences between the two. Where δ1=0.17, δ2=0.09;

[0089]

[0090] Where δ represents the standard deviation of the original image pixels, δ1 = 0.17, δ2 = 0.09;

[0091] 5) This invention applies the standard deviation δ of pixels to different L... ave Different adjustment functions are used to enhance the curvature of the global tone mapping function curve, thereby improving its adaptability to various dark scenes with varying brightness levels. The final pixel enhancement curve changes are shown in the figure below, where the three different curves represent different L values. ave The relationship between pixel value and enhancement value under certain conditions (L corresponding to the blue curve) ave Maximum, L corresponding to the orange curve ave Secondly, the L corresponding to the green curve ave (Minimum). As can be seen, as the logarithmic mean of luminance decreases, the enhancement ratio of global tone mapping for darker pixels gradually increases, while the enhancement ratio for brighter pixels remains relatively constant. Therefore, the logarithmic mean of luminance can be adjusted to achieve a better enhancement effect.

[0092] 6) The optimization strategy for the logarithmic mean of brightness can be represented by the following piecewise function, where Y c This represents the logarithmic mean of the brightness after adjustment by the adjustment function, where δ1 = 0.17.

[0093] δ2 = 0.09;

[0094]

[0095] 7) Based on the above formula, we obtain Y c To improve the algorithm's ability to enhance darker areas and suppress brighter areas, a parameter P1 is introduced to adjust Y. c and V max The adjusted formula is shown below:

[0096]

[0097] Where Y cln V represents the logarithmic mean of the brightness obtained after the above adjustments. maxln P1 represents the maximum brightness obtained after the above adjustments, and P1 represents an empirical parameter used to adjust the representation ability of the enhancement curve to improve the final effect of the present invention. In the present invention, P1 is set to 13.

[0098] 8) According to the Weber-Fechner theorem, the human eye's perception of an image closely matches the characteristics of a log function. Therefore, the Y obtained in the previous steps can be used... cln and V maxln Substitute the values ​​into the following formula for calculation, and introduce the adjustment parameter P2, where L e This indicates the initial enhancement result obtained after global tone mapping in this invention;

[0099]

[0100] 9) In this invention, the parameters P1 and P2 are generally set to 10 and 0.85, respectively. Furthermore, based on the previous step, this invention uses Newton's iterative algorithm to compress the global tone enhancement curve to further improve algorithm performance. Specifically, this invention uses Newton's iterative algorithm to calculate an enhancement curve cutoff threshold V that approximates the threshold obtained due to the parameters P1 and P2. cut In this invention, the loop parameter of the Newton iteration method is set to 20. Only pixels smaller than this threshold will be enhanced by the enhancement curve. Therefore, we can obtain:

[0101]

[0102] Specifically, the gradient weight local enhancement module:

[0103] 1) A low-light image often has low contrast, and many valuable details in the image are difficult to recover after global tone mapping. However, gradient information can well characterize the changes in brightness and darkness, edge contours, and other information in the image. This invention uses the Scharr operator to convolve the brightness component V obtained in the global tone mapping module, thereby extracting the gradient information at each pixel and obtaining gradient weight information maps in the x and y directions.

[0104] 2) The Scharr operator has a size of 3×3. It amplifies the differences between pixels through the weight coefficients in the filter. The Scharr in the x and y directions is defined as follows:

[0105]

[0106] 3) Perform a convolution operation on the brightness component V using the two Scharr operators from the previous step, and obtain two matrices g of the same size as the original image. x and g y The specific calculation formula is shown below, where G x () denotes a convolution operation on an image, where the convolution kernel is G. x G y () denotes a convolution operation on an image, where the convolution kernel is G. y ;

[0107] g x (x,y)=G x (V(x,y)),g y (x,y)=G y (V(x,y)) (12)

[0108] 4) In an image, areas with significant brightness variations and object edges typically have larger gradients, while areas with less pronounced brightness variations and non-edge areas usually have smaller gradients. Considering that image enhancement should focus on areas with less pronounced brightness variations and non-edge areas, the gradient enhancement weight map is obtained using the following formula, where σ represents...

[0109] A very small constant is used to avoid the gradient being zero.

[0110]

[0111] 5) The final enhanced weight map is obtained by combining the luminance component V with the re weight map obtained in the previous step and performing a power law transformation, where β = 255, which is used to normalize the luminance component V to the range (0,1), and α is an empirical parameter, which is set to 0.5 in this invention.

[0112]

[0113] 6) The gradient detection local enhancement module is used to further enhance the image after global tone mapping, which can restore image detail information. The image after local enhancement is obtained by the following formula:

[0114]

[0115] Specifically, the glare suppression module:

[0116] 1) When performing low-light enhancement, existing algorithms often inappropriately enhance the brighter parts of the image (such as car headlights and streetlights), making the surrounding areas too bright and resulting in a poor visual experience. This invention employs a linear extension algorithm for glare suppression;

[0117] 2) Weber's theorem states that the difference in light intensity that people can perceive is related to the background light intensity.

[0118] Regions with large changes in amplitude are more likely to attract the attention of the human eye. Based on this characteristic, when processing images, low-gray-level and high-gray-level regions with narrow amplitudes can be stretched, while intermediate-gray-level regions can be compressed. This invention employs a linear expansion algorithm that matches the human eye, as shown in the following formula, where L... e This indicates that it has been enhanced by the above two modules.

[0119] The image after L o This represents the original image, and a1 and a2 represent the corresponding weights.

[0120] L p =a1L o (x,y)+a2L e (x,y) (16)

[0121] L p This represents the result obtained after applying the linear extension algorithm;

[0122] 3) The weights a1 and a2 are calculated as follows. To achieve the desired effect, the weights of darker pixels should be increased in brighter scenes and decreased in darker scenes. The threshold T = 256 - V. ave a2 = 1 - a1, where V ave This represents the average brightness of the image. This method can adaptively calculate the threshold for different images, improving the flexibility of the linear expansion algorithm and enabling the image to achieve better visual effects.

[0123]

[0124] 4) After the image is processed by the glare suppression module, brighter areas can be suppressed, but a certain degree of contrast reduction and unnatural colors may occur. Since the basic principle of the CLAHE algorithm is to divide the image into multiple sub-blocks and perform histogram equalization with limited contrast, it mainly considers the local information of the image and will not affect the enhancement effect in darker areas due to uneven image lighting. Therefore, the CLAHE algorithm is used to further improve image contrast while preserving the glare suppression effect.

[0125] 5) In this invention, the CLAHE algorithm is not directly applied to the three RGB channels of the enhanced image. Instead, the image is first converted from RGB space to XYZ space, then from XYZ space to Lab space, and then the CLAHE algorithm is applied to the L component. The formula used for space conversion is as shown in (18).

[0126] As shown in (19) and (20):

[0127]

[0128]

[0129] 6) Apply the CLAHE algorithm to the L component obtained in the previous step. In this invention, the contrast l in the CLAHE algorithm is set to 1, and the size of the divided sub-images is set to 10×10. After enhancement by the CLAHE algorithm, the final output image is obtained, as shown below, where L... Le This represents the L component after Le is transformed to Lab space;

[0130] L out =CLAHE(L Le (21)

[0131] 7) Combine the L component obtained after contrast processing of the image with the other unprocessed components and convert them back to the RGB space to obtain the final enhanced image of the present invention.

[0132] ①The advantages and uses of this product.

[0133] like Figure 4 As shown, compared to existing low-light image enhancement methods, this invention combines low-light image enhancement with glare suppression, taking into account various nighttime scenes and lighting conditions. It not only effectively recovers details in darker areas of the image but also avoids amplifying existing bright spots and halos, suppressing brighter areas to a certain extent and achieving overall image brightness balance. Furthermore, this invention achieves a good balance between enhancement effect and efficiency, maintaining good real-time performance across various image resolutions. In summary, this invention can simultaneously handle restoration tasks in various lighting scenarios, providing good enhancement effects while ensuring good real-time performance for practical applications.

[0134] ②The process or method of using the product.

[0135] This invention, as an image processing algorithm, can be combined with existing devices and deployed on their computing backend to perform secondary processing and enhancement on images. Specifically, it can be used for:

[0136] Reconnaissance drones: These drones can be used for reconnaissance, ensuring image visibility in dark environments and improving reconnaissance efficiency and effectiveness.

[0137] Intelligent vehicles: Can be used in intelligent autonomous vehicles, enabling autonomous driving tasks at night or in low-light conditions, ensuring the accuracy and safety of autonomous driving.

[0138] Individual soldier equipment: Can be used for individual combat equipment, enhancing soldiers' night combat capabilities, reducing dependence on light sources and ensuring stealth.

[0139] Special vehicles: Can be used for special vehicles to ensure that they can perform corresponding tasks at night and in low light conditions, thereby improving efficiency and response speed.

[0140] Security monitoring: Can be used in security monitoring cameras, enabling video surveillance in various situations, improving image detail and visual effects.

Claims

1. A glare suppression low-light image enhancement method based on global mapping and local tuning, characterized in that, It comprises the following steps: Step S1: the original image is preliminarily enhanced, global tone mapping is performed through an improved algorithm parameter optimization strategy to restore some details and information of the original image, and the color saturation and contrast of the image are adjusted; Step S2: the image region gradient detection is performed on the original image to calculate the region with relatively sharp light and dark changes, and the image is adaptively enhanced through the corresponding algorithm; Step S3: the linear expansion algorithm is used to superimpose the images before and after enhancement through the pixel-level weight adaptive algorithm, so that the glare in the image is suppressed, and the contrast and color of the image after glare suppression are adaptively adjusted by the algorithm, so that the image conforms to the habit of human eye vision; The step S2 comprises: Step S21: the Scharr operator is used to perform convolution operation on the luminance component V obtained in the global tone mapping module, so as to extract the gradient information at each point pixel and obtain the gradient weight information graph in x direction and y direction; Step S22: the size of the Scharr operator is 3*3, the difference between pixels is enlarged through the weight coefficient in the filter, and the Scharr in x direction and y direction is defined as follows: (11) Step S23: Perform a convolution operation on the brightness component V using the two Scharr operators from step S22, and obtain two matrices g of the same size as the original image. x and g y The specific calculation formula is shown below, where G x ( ) represents a convolution operation on an image, and the convolution kernel is G. x G y ( ) represents a convolution operation on an image, and the convolution kernel is G. y ; (12); Step S24: Obtain the enhanced weight map of gradient information according to the following formula, wherein represents a very small constant to avoid the case that the gradient is 0. (13); Step S25: the final enhancement weight graph is obtained by combining the luminance component V and the re weight graph obtained in the previous step and performing a power law transformation, wherein β=255, the function is to normalize the luminance component V to (0, 1), and α is an empirical parameter; (14) Step S26: the gradient detection local enhancement module is used to enhance the image after global tone mapping, and the image details are recovered, and the image after local enhancement is obtained by the following formula: (15) The step S3 specifically comprises: Step S31: the linear expansion algorithm is used for glare suppression; Step S32: a linear expansion algorithm matching human eyes is adopted, and the algorithm is specifically as follows: L e represents the image after enhancement, L o represents the original image, and represents the corresponding weight; (16) L p represents the result obtained after the linear expansion algorithm; Step S33: weight in linear expansion algorithm and Different values are taken at different brightness positions. In the brighter pixel area, the weight of the original image is higher, and in the darker pixel area, the weight of the original image is lower; wherein the threshold value , wherein V ave represents the average value of the image brightness; (17) Step S34: the CLAHE algorithm is used to improve the image contrast on the basis of retaining the glare suppression effect; Step S35: the image is converted from RGB space to XYZ space, and then from XYZ space to Lab space, and then the CLAHE algorithm is applied to the L component, and the formula applied in the space conversion is shown in (18), (19) and (20): (18) (19) (20)。 2. The glare-inhibited low-illumination image enhancement method based on global mapping and local optimization according to claim 1, characterized in that, The step S1 specifically comprises the following steps: Step S11: the specific calculation formula of converting the original image from RGB space to HSV space is as follows: (1) (2) (3) Wherein, H represents the hue component of HSV space, which refers to the color type or color phase, S represents the saturation component of HSV space, which refers to the color brightness, and V represents the brightness component of HSV space, which refers to the color brightness; Step S12: Calculate the logarithmic average value Y of the brightness using the V component calculated in step Sll ave : (4) where MN represents the product of the height and width of the original image, where C represents a constant, x represents the horizontal coordinate of a pixel point, and y represents the vertical coordinate of the pixel point. Step S13: Calculate the standard deviation of all pixels in the V component using formula (5) where v(i,j) represents a pixel on the V component, represents the mean value of pixels of the original image; (5) Step S14: according to different pixel standard deviation, the input original image is classified according to formula (6), and the input original image is classified into three categories: non-uniformly illuminated image, uniformly illuminated image and image with light and dark difference between the two; (6)； wherein denotes the input raw image pixel standard deviation, ; Step S15: adjusting the value of the standard deviation of the pixels ave The different adjustment functions are used to enhance the curvature of the global tone mapping function curve, so as to enhance the adaptability to different types of dark scenes with different brightness conditions.​ 3. The glare-inhibited low-illumination image enhancement method based on global mapping and local optimization according to claim 2, characterized in that, The step S1 further comprises: Step S16: The optimization adjustment strategy of the luminance logarithmic mean is specifically shown by the following piecewise function, wherein Y c represents the luminance logarithmic mean after the adjustment function adjustment. (7) Step S17: Introducing parameter P1 to adjust Y c and V max The adjustment formula is as follows: (8) where Y cln represents the adjusted luminance log mean, V maxln represents the adjusted maximum luminance, and P1 represents an empirical parameter used to adjust the representability of the enhancement curve to improve the final result. Step S18: using the obtained Y cln and V maxln Bringing into the following formula for calculation, while introducing the adjustment parameter P2 in it, where L e represents the result obtained after the preliminary enhancement of global tone mapping; (9) Step S19: Calculate the threshold value V of the enhancement curve cut-off approaching the enhancement curve due to the P1 and P2 parameters using the Newton iteration algorithm cut where the loop parameter of the Newton iteration method is set to 20, and only the pixel part less than the threshold value is enhanced by the enhancement curve, thus obtaining: (10)。 4. The glare-inhibited low-illumination image enhancement method based on global mapping and local optimization according to claim 1, characterized in that, The step S3 specifically further comprises: Step S36: applying the CLAHE algorithm to the L component obtained in step S35, setting the contrast l in the CLAHE algorithm to 1, and setting the divided sub-image size size to 10x10, and after the CLAHE algorithm enhancement, the final output image is obtained, as follows, wherein L Le represents the L component after Le conversion to the Lab space; (21) Step S37: the L component obtained after contrast processing of the image and the other components without processing are combined and converted back to RGB space to obtain the final enhanced image.

5. A glare-inhibited low-illumination image enhancement system based on global mapping and local tuning, characterized in that, The low-illumination image enhancement method based on global mapping and local optimization for glare suppression comprises a global tone mapping enhancement module, a gradient detection local enhancement module, and a glare suppression module.

6. A glare-inhibited low-illumination image enhancement system based on global mapping and local tuning according to claim 5, characterized in that, The global tone mapping enhancement module is responsible for the preliminary enhancement of the original image, and the global tone mapping is performed through an improved algorithm parameter optimization strategy to make the original image restore certain details and information, and adjust the color saturation and contrast of the image.

7. The glare-inhibited low-light image enhancement system based on global mapping and local tuning of claim 5, wherein, The gradient detection local enhancement module calculates the regions with sharp light and dark changes by performing image region gradient detection on the original image, and adaptively enhances the image through a corresponding algorithm.

8. The glare-inhibited low-light image enhancement system based on global mapping and local tuning of claim 5, wherein, The glare suppression module suppresses the glare in the image by a linear expansion algorithm, superimposes the images before and after enhancement through a pixel-level weight adaptive algorithm, and adaptively adjusts the contrast and color of the image after glare suppression through an algorithm, so that the image conforms to the habit of human eye vision.

Images