A method, system, device and storage medium for detail enhancement of low-light images

Abstract

The application discloses a low-illumination image detail enhancement method, system, device and storage medium, and belongs to the technical field of image processing.The application converts an image from an RGB space to a YCbCr color space, separates a Y component representing overall brightness information of the image, so that the interference of chroma information of other channels on data processing is reduced in a subsequent processing process; meanwhile, the application can adjust appropriate gain parameters and fusion parameters according to specific illumination characteristics and scenes, so that the contrast and brightness of the low-illumination image are improved as a whole, the detail part of the low-illumination image is enhanced, and the picture quality of the low-illumination image is improved; finally, a linear mapping reconstruction technology is used to reconstruct the RGB color space, the color distortion problem possibly occurring in the reconstructed color space is improved, and the overall quality of the image is improved.

Description

Technical Field

[0001] This invention belongs to the field of image processing technology, and specifically relates to a method, system, device, and storage medium for enhancing details in low-light images. Background Technology

[0002] Low-light conditions refer to the illumination level of the environment in which the subject is located, which is below a certain value, causing the video signal level of the detector to also fall below a certain value. Low-light images are characterized by generally low grayscale values, an overall dark appearance, significant image noise, relatively concentrated pixel distribution, small grayscale differences between edge pixels and scene pixels, and insufficient detail. Low-light image detail enhancement technology is one of the important directions in the development of image processing technology both domestically and internationally. This technology improves the ability to detect targets and analyze scenes in low-light environments.

[0003] Traditional low-light image detail enhancement techniques adjust the overall pixels of the image. However, the enhancement results in problems such as local halos, overexposure of highlights, and loss of details due to insufficient brightness in shadows. In practical applications, it is difficult to guarantee adaptability to various low-light scenes. Furthermore, the enhancement process introduces new image noise, which is not conducive to the subsequent processing and analysis of image information. Summary of the Invention

[0004] To address the problem of poor detail enhancement in existing low-light images, this invention provides a method, system, device, and storage medium for enhancing the details of low-light images. This invention improves the overall contrast of low-light images, enhances image details, and effectively suppresses noise.

[0005] This invention is achieved through the following technical solution:

[0006] A method for enhancing details in low-light images, the method comprising:

[0007] Acquire low-light image data to obtain the original low-light RGB image;

[0008] The original low-light RGB image is converted from the RGB space to the YCbCr color space, and the Y channel grayscale image representing the image brightness is separated.

[0009] The grayscale image of the Y channel is used as a gradient-guided image, and guided filtering is applied to it to obtain a smooth and denoised image that retains image details.

[0010] A sigmoid function with an adjustable gain factor is used to perform brightness stretching on the grayscale image of the Y channel, and a Gaussian filter is applied to the processed image to obtain a linearly enhanced brightness image.

[0011] The smoothed and denoised image and the linearly enhanced brightness image are subjected to gray-level difference processing to obtain a detail extraction image that contains only image details;

[0012] The brightness linearly enhanced image and the detail extraction image are subjected to image fusion processing, and the fusion coefficient is adjusted using the fitness coefficient to obtain a fused detail-enhanced image in which both details and brightness are enhanced.

[0013] The fused detail-enhanced image is remapped in RGB color space using a linear mapping method based on the color information of the original low-light RGB image to obtain a detail-enhanced low-light RGB image.

[0014] This invention converts images from the RGB color space to the YCbCr color space and separates the Y component, which represents the overall brightness information of the image, to reduce interference from the chromaticity information of other channels in subsequent processing. Simultaneously, it employs a sigmoid function with an adjustable gain factor for brightness stretching to improve brightness and contrast. Compared to existing techniques that use fixed processing parameters to achieve brightness or contrast gain, this invention can adjust appropriate gain and fusion parameters according to specific lighting characteristics and scenes, resulting in an overall improvement in the contrast and brightness of low-light images while enhancing details and improving image quality. Finally, linear mapping reconstruction technology is used to reconstruct the RGB color space, mitigating potential color distortion issues in the reconstructed color space and further improving overall image quality.

[0015] In a preferred embodiment, the present invention employs a guided filtering function to perform guided filtering processing on the gradient guided image, wherein the guided filtering function is:

[0016]

[0017] In the formula, p is the input image, q is the output image, I is the gradient-guided image, and |ω| represents ω. k The number of pixels in the window, μ k and σ 2 This represents the value and variance of the gradient-guided plot within the high window, where ε is the smoothing factor.

[0018] The local linear relationship between the output image and the gradient-guided image is expressed as follows:

[0019]

[0020] In the formula, α k With β k These are the linear coefficients, ω k In the gradient-guided image I, a template window is defined with pixel k as its center and radius N, and within the window α...k With β k It is a constant value.

[0021] In a preferred embodiment, the adjustable sigmoid function for the gain factor of the present invention is:

[0022]

[0023] In the formula, Imax and Imin correspond to the maximum and minimum gray values ​​of the original image, respectively; I is the pixel gray value of the input image; a is the parameter controlling the intensity of the linear gradient; and b is the parameter controlling the range of the linear gradient intensity.

[0024] In a preferred embodiment, the present invention performs Gaussian filtering on the image using the following formula:

[0025]

[0026] In the formula, x 2 With y 2 represents the distance between the remaining pixels in the neighborhood and the center pixel in the neighborhood, respectively; c represents the standard deviation.

[0027] In a preferred embodiment, the present invention uses the following formula to perform weighted fusion processing on the brightness linear enhancement image and the detail extraction image:

[0028] I5 = w·I3 + (1-w)I4

[0029] In the formula, I5 is the fused detail-enhanced image; w is the weight coefficient of the linear brightness enhancement image I3; and 1-w is the weight coefficient of the detail extraction image I4.

[0030] As a preferred embodiment, the present invention adjusts the fusion coefficient using the following formula:

[0031] A = max∑δ(x,y) 2 P δ (x,y)

[0032] In the formula, δ(x,y) is the gray-level difference between adjacent pixels; P δ (x,y) represents the pixel distribution probability with a gray-level difference of δ between adjacent pixels.

[0033] In a preferred embodiment, the linear mapping coefficients of the present invention are expressed as follows:

[0034]

[0035] In the formula, I0 is the original low-light RGB image; I5 is the fused image with enhanced details.

[0036] The mapping relationship is represented as follows:

[0037]

[0038] In the formula, [R0,G0,B0] are the RGB color channels of the original low-light image, and [R',G',B'] are the RGB color channels after color remapping.

[0039] Secondly, this invention proposes a detail enhancement system for low-light images, comprising:

[0040] The image acquisition module collects low-light image data to obtain the original low-light RGB image;

[0041] The color space conversion module converts the original low-light RGB image from the RGB space to the YCbCr color space and separates the Y channel grayscale image representing the image brightness.

[0042] The guided filtering module uses the Y-channel grayscale image as a gradient guided image and performs guided filtering on it to obtain a smooth and denoised image that retains image details.

[0043] The linear enhancement module uses a sigmoid function with an adjustable gain factor to perform brightness stretching on the grayscale image of the Y channel, and then performs Gaussian filtering on the processed image to obtain a linearly enhanced brightness image.

[0044] The grayscale difference module performs grayscale difference processing on the smoothed and denoised image and the brightness linearly enhanced image to obtain a detail extraction image that contains only image details;

[0045] The fusion module performs image fusion processing on the brightness linearly enhanced image and the detail extraction image, and adjusts the fusion coefficient using an fitness coefficient to obtain a fused detail-enhanced image with both enhanced detail and brightness.

[0046] The color space remapping module uses a linear mapping method based on the color information of the original low-light RGB image to perform RGB color space remapping on the fused detail-enhanced image, thereby obtaining a detail-enhanced low-light RGB image.

[0047] Thirdly, the present invention provides a computer device including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of the method described above.

[0048] Fourthly, the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the method described above.

[0049] Compared with the prior art, the present invention has the following advantages and beneficial effects:

[0050] 1. This invention uses a sigmoid function with an adjustable gain factor to perform brightness stretching processing to improve its brightness and contrast. For low-light images with different image features or grayscale histogram features, the coefficients can be adjusted according to their characteristics to optimize local processing and make the final gain effect more ideal. This invention has higher adaptability and flexibility to different low-light environments.

[0051] 2. This invention evaluates the fusion coefficient using the fitness coefficient and establishes a mapping relationship between the fitness coefficient and the fusion weight, making the fusion results more intuitive and reliable.

[0052] 3. This invention employs a linear mapping reconstruction method, which differs from traditional RGB color space reconstruction. This method reduces the computational load of the algorithm and can better improve the color distortion problem that may occur in the reconstructed color space, thereby improving the overall image quality and restoring the colors of low-light images. Attached Figure Description

[0053] The accompanying drawings, which are included to provide a further understanding of embodiments of the invention and form part of this application, do not constitute a limitation thereof. In the drawings:

[0054] Figure 1 This is a schematic diagram of the method flow according to an embodiment of the present invention;

[0055] Figure 2 This is a system principle block diagram according to an embodiment of the present invention;

[0056] Figure 3 The images shown are comparisons of low-light images before and after processing using the method of this embodiment of the invention; where (a) is the image before processing and (b) is the image after processing. Detailed Implementation

[0057] To make the objectives, technical solutions, and advantages of the present invention clearer, the present invention will be further described in detail below with reference to the embodiments and accompanying drawings. The illustrative embodiments and descriptions of the present invention are only used to explain the present invention and are not intended to limit the present invention.

[0058] Example 1

[0059] To achieve effective enhancement of low-light images, this embodiment proposes a detail enhancement method for low-light images to solve the problem of poor detail enhancement effect in existing low-light images.

[0060] Specifically, such as Figure 1 As shown, the method proposed in this embodiment specifically includes the following steps:

[0061] Step 1: Collect low-light image data containing valid information to obtain the original low-light RGB image I0;

[0062] Step 2: Convert the original low-light RGB image I0 from RGB space to YCbCr color space, and separate the Y channel grayscale image I1 representing the image brightness. In this embodiment, the image is mapped from RGB to YCbCr space, and the Y channel grayscale image representing the image brightness information is extracted separately. This can effectively avoid the interference that the color information of other channels may cause in subsequent processing, and at the same time reduce the amount of computation of the algorithm.

[0063] Step 3: Use the Y channel grayscale image I1 as the gradient guide image and apply guided filtering to it to obtain a smooth and denoised image I2 that retains the image details.

[0064] Step 4: The brightness of the Y channel grayscale image I1 is stretched using the Sigmoid function with adjustable gain factor to improve its brightness and contrast. The processed image is then subjected to Gaussian filtering to obtain the linearly enhanced brightness image I3.

[0065] Step 5: Perform gray-level difference processing on the smoothed and denoised image I2 and the brightness linearly enhanced image I3 to obtain the detail extraction image I4 containing only image details;

[0066] Step 6: The brightness linear enhancement image I3 and the detail extraction image I4 are fused using a weighted fusion strategy, and the fusion coefficient is adjusted using the fitness coefficient to obtain the fused detail enhancement image I5 with the added detail and brightness enhancement.

[0067] Step 7: Perform RGB color space remapping on the fused detail-enhanced image I5. Use a linear mapping method based on the color information of the original low-light RGB image I0 to achieve RGB color space remapping of image I5, and obtain the final detail-enhanced low-light RGB image I'.

[0068] In an optional implementation, step 3 further includes:

[0069] Let the input image be p (i.e., the grayscale image I1 of the Y channel), the output image be q (the smoothed and denoised image I2), and the image used for gradient guidance be I (i.e., the grayscale image I1 of the Y channel). The local linear relationship between the output image q and the gradient-guided image is shown in the following equation:

[0070]

[0071] In the formula, α k With β k These are the linear coefficients, ω kIn the gradient-guided image I, a template window is defined with pixel k as its center and radius N, and within the window α... k With β k This is a constant value. The purpose of this window template is to maintain pixel gradient consistency in edge details between the output image and the gradient-guided image, achieving filtering that preserves details while smoothing noise. The following constraints are used to minimize the difference between the output image q and the input image p.

[0072]

[0073] Where ε is the regularization parameter, and α is obtained by solving this function. k With β k They are respectively:

[0074]

[0075]

[0076] in:

[0077]

[0078] In the formula, |ω| represents ω k The number of pixels in the window, μ k and σ 2 This represents the value and variance of the gradient-guided plot within the high window, where ε is the smoothing factor.

[0079] α k and β k Bring back to q i The specific formula for the filtering process can be obtained from the calculation formula:

[0080]

[0081] The above explains the derivation principle of gradient-guided filtering. When using gradient-guided filtering to enhance image details, it maintains the consistency of pixel gradients between the input and output images, thus ensuring that the retained details are those of the original image. As the derivation shows, gradient-guided filtering can effectively preserve edge details, performing smoothing filtering only on flat areas of the image. However, when using local variance to determine image edge details, since the smoothing factor is a fixed parameter, some local details may be filtered out due to their small variance, although this ensures that most noise is eliminated.

[0082] In one optional implementation, step 4 uses a sigmoid function with an adjustable gain factor to perform brightness stretching on the grayscale image I1, the function of which is as follows:

[0083]

[0084] Where Imax and Imin correspond to the maximum and minimum grayscale values ​​of the original image, respectively; I is the pixel grayscale value of the input image; a is a parameter controlling the intensity of the linear gradient, and the larger the value, the stronger the gain function gain; b is a parameter controlling the range of the linear gradient intensity, and a smaller b results in a more significant contrast gain in dark areas, while a larger b results in a more significant contrast gain in bright areas. By selecting appropriate values ​​for the a and b parameters, the image can be adjusted to achieve the desired effect, thereby improving its brightness and contrast. For example, for low-light images with low overall contrast, the values ​​of a and b need to be selected to make the gain curve enhance the entire image relatively evenly. When local brightness is insufficient, such as excessively dark shadows or areas with high contrast, the value of b needs to be reduced to focus on improving the brightness and contrast of the dark areas. When local glare appears in the image, the value of b needs to be increased to reduce the impact of glare on the image.

[0085] In one optional implementation, step 4 involves applying a Gaussian filter to the image after brightness stretching to improve contrast, using the following formula:

[0086]

[0087] Where, x 2 With y 2 represents the distance between the remaining pixels in the neighborhood and the center pixel in the neighborhood, respectively; c represents the standard deviation. The larger the standard deviation, the wider the two-dimensional Gaussian image, and the more obvious the filtering effect on noise.

[0088] In one optional implementation, step 6 uses the following formula to perform image fusion on the brightness linear enhancement image I3 and the detail extraction image I4 using a weighted fusion strategy:

[0089] I5 = w·I3 + (1-w)I4

[0090] Where w represents the weighting coefficient of the linearly enhanced brightness image I3; and 1-w represents the weighting coefficient of the detail extraction image I4. The fitness coefficients are adjusted using a coefficient fitness function, the calculation formula of which is as follows:

[0091] A = max∑δ(x,y) 2 P δ (x,y)

[0092] Where δ(x,y) is the gray-level difference between adjacent pixels; P δ(x,y) represents the pixel distribution probability with a gray-level difference of δ between adjacent pixels. The fitness coefficient A is the result of calculating the pixel values ​​of the fusion result. Different fusion weights w have a one-to-one mapping relationship with the fitness coefficient A. Therefore, the fitness coefficient A can describe the overall effect of fusion and can form a fixed evaluation index (which can be formed into a table or fitted curve by changing the value of w multiple times).

[0093] In an optional implementation, step 7 employs a linear mapping method based on the color information of the original low-light RGB image I0 to achieve RGB color remapping in order to reduce color distortion. The linear mapping coefficients are expressed as follows:

[0094]

[0095] Where I0 is the original low-light RGB image; I5 is the fused, detail-enhanced image; let [R0,G0,B0] be the RGB color channels of the original low-light image, and [R',G',B'] be the RGB color channels after color remapping, with the mapping relationship as follows:

[0096]

[0097] Traditional YCbCr to RGB color space conversion generally uses a fixed conversion calculation method, which may result in significant color distortion after restoring the RGB color space, and the algorithm has high time complexity. In the RGB space, if there are two pixels with a proportional ratio, these two points have the same color, only different brightness information. By using the ratio of these two points, a scaling factor L can be obtained. By processing all pixels with this scaling factor, the problem of color distortion can be effectively avoided. Therefore, this embodiment adopts the linear mapping method.

[0098] This embodiment also proposes a detail enhancement system for low-light images, specifically as follows: Figure 2 As shown, the system includes:

[0099] The image acquisition module collects low-light image data containing valid information to obtain the original low-light RGB image I0;

[0100] The color space conversion module converts the original low-light RGB image I0 from the RBG space to the YCbCr color space and separates the Y channel grayscale image I1, which represents the image brightness.

[0101] The guided filtering module uses the grayscale image I1 of the Y channel as the gradient guide image and processes it with guided filtering to obtain a smooth and denoised image I2 that retains the image details.

[0102] The linear enhancement module uses a sigmoid function with an adjustable gain factor to perform brightness stretching on the Y channel grayscale image I1 to improve its brightness and contrast, and then performs Gaussian filtering on the processed image to obtain the linearly enhanced brightness image I3.

[0103] The grayscale difference module performs grayscale difference processing on the smoothed and denoised image I2 and the brightness linearly enhanced image I3 to obtain a detail extraction image I4 that contains only image details;

[0104] The fusion module performs image fusion using a weighted fusion strategy on the brightness linear enhancement image I3 and the detail extraction image I4, and adjusts the fusion coefficient using an fitness coefficient to obtain the fused detail-enhanced image I5 with both detail and brightness enhancement.

[0105] The color space remapping module performs RGB color space remapping on the fused detail-enhanced image I5. It uses a linear mapping method based on the color information of the original low-light RGB image I0 to achieve RGB color space remapping of image I5, resulting in the final detail-enhanced low-light RGB image I'.

[0106] This embodiment also proposes a computer device for performing the methods described above in this embodiment.

[0107] Computer devices include a processor, internal memory, and a system bus; various device components, including internal memory and the processor, are connected to the system bus. A processor is hardware used to execute computer program instructions through basic arithmetic and logical operations within the computer system. Internal memory is a physical device used for temporary or permanent storage of computational programs or data (e.g., program state information). The system bus can be any of several types of bus architectures, including a memory bus or memory controller, a peripheral bus, and a local bus. The processor and internal memory can communicate via the system bus. Internal memory includes read-only memory (ROM) or flash memory, and random access memory (RAM), which typically refers to the main memory loaded with the operating system and computer programs.

[0108] Computer devices typically include an external storage device. The external storage device can be selected from a variety of computer-readable media, which are any usable media accessible by a computer device, including both removable and fixed media. Examples of computer-readable media include, but are not limited to, flash memory (microSD cards), CD-ROMs, digital versatile optical discs (DVDs) or other optical disc storage, magnetic tape cartridges, magnetic tapes, disk storage or other magnetic storage devices, or any other media that can be used to store desired information and is accessible by a computer device.

[0109] Computer devices can logically connect to one or more network terminals in a network environment. Network terminals can be personal computers, servers, routers, smartphones, tablets, or other public network nodes. Computer devices connect to network terminals through network interfaces (LAN interfaces). A Local Area Network (LAN) is a computer network interconnected within a limited area, such as a home, school, computer lab, or office building using network media. WiFi and twisted-pair Ethernet are the two most commonly used technologies for building LANs.

[0110] It should be noted that other computer systems, including more or fewer subsystems than computer equipment, are also applicable to the invention.

[0111] As described in detail above, the computer device applicable to this embodiment can perform specified operations of the multidimensional spatial horizon search method. The computer device performs these operations through software instructions executed by a processor in a computer-readable medium. These software instructions can be read into memory from a storage device or from another device via a local area network interface. The software instructions stored in memory cause the processor to execute the aforementioned method for processing group membership information. Furthermore, the present invention can also be implemented through hardware circuitry or hardware circuitry combined with software instructions. Therefore, implementation of this embodiment is not limited to any specific combination of hardware circuitry and software.

[0112] Example 2

[0113] This embodiment uses the low-light image detail enhancement method proposed in Embodiment 1 above to enhance a certain acquired low-light original image data (such as...). Figure 3 The process shown in (a) is as follows:

[0114] (1) Convert the original low-light image I0 from RGB space to YCbCr color space, and separate a Y channel grayscale image I1 representing the image brightness. The linear conversion formula from RGB to YCbCr color space is as follows:

[0115]

[0116] (2) The grayscale image I1 of the Y channel is used as the gradient-guided image. The gradient-guided filtering formula of the above embodiment 1 is used to process it with guided filtering to obtain a smooth and denoised image I2 that retains the image details.

[0117] (3) The grayscale image I1 of the Y channel is subjected to brightness stretching processing using the Sigmoid function F(I) with adjustable gain factor in Example 1 to improve its brightness and contrast, and the processed image is subjected to Gaussian filtering using the Gaussian filtering function G(x,y) in Example 1 to obtain a brightness linearly enhanced image I3.

[0118] (4) Using the gray-level difference formula of Example 1, the smoothed and denoised image I2 and the brightness linearly enhanced image I3 are subjected to gray-level difference to obtain a detail extraction image I4 containing only image details.

[0119] (5) The weighted fusion calculation formula is used to perform weighted fusion calculation on the brightness linear enhancement image I3 and the detail extraction image I4, and the coefficient adaptation function is used to adjust the fusion coefficient to obtain a fused detail enhancement image I5 with enhanced details and brightness.

[0120] (6) The color space is remapped on the fused detail-enhanced image using the linear mapping relationship of Example 1 to obtain the final detail-enhanced low-light RGB image I', as shown below. Figure 3 As shown in (b) of the diagram.

[0121] Depend on Figure 3 As can be seen, after processing the acquired low-light original image data using the low-light image detail enhancement method proposed in Example 1, the overall brightness and contrast of the low-light image are improved, the overall image brightness is increased, the details that were originally in the dark area are enhanced, and no new noise is introduced significantly.

[0122] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for enhancing details in low-light images, characterized in that, The method includes: Acquire low-light image data to obtain the original low-light RGB image; The original low-light RGB image is converted from the RGB space to the YCbCr color space, and the Y channel grayscale image representing the image brightness is separated. The grayscale image of the Y channel is used as a gradient-guided image, and guided filtering is applied to it to obtain a smooth and denoised image that retains image details. A sigmoid function with an adjustable gain factor is used to perform brightness stretching on the grayscale image of the Y channel, and a Gaussian filter is applied to the brightness-stretched image to obtain a linearly enhanced brightness image. The smoothed and denoised image and the linearly enhanced brightness image are subjected to gray-level difference processing to obtain a detail extraction image that contains only image details; The brightness linear enhancement image and the detail extraction image are subjected to image fusion processing, and the fusion coefficient is adjusted by using a fitness coefficient based on the gray level difference between adjacent pixels and their corresponding pixel distribution probability to obtain a fused detail-enhanced image with enhanced detail and brightness. The fused detail-enhanced image is remapped in RGB color space using a linear mapping method based on the color information of the original low-light RGB image to obtain a detail-enhanced low-light RGB image. The adjustable Sigmoid function for the gain factor is: ； In the formula, Imax and Imin correspond to the maximum and minimum gray values ​​of the original image, respectively; I is the pixel gray value of the input image; a is the parameter controlling the intensity of the linear gradient; and b is the parameter controlling the range of the linear gradient intensity.

2. The method for enhancing details in low-light images according to claim 1, characterized in that, A guided filtering function is used to perform guided filtering on the gradient guided image. The guided filtering function is as follows: ； In the formula, p is the input image, q is the output image, I is the gradient-guided image, and |ω| represents ω. k The number of pixels in the window, μ k and σ 2 This represents the value and variance of the gradient-guided plot within the high window, where ε is the smoothing factor. The local linear relationship between the output image and the gradient-guided image is expressed as follows: ； In the formula, α k With β k These are the linear coefficients, ω k In the gradient-guided image I, a template window is defined with pixel k as its center and radius N, and within the window α... k With β k It is a constant value.

3. The method for enhancing details in low-light images according to claim 1, characterized in that, The image is processed using the following Gaussian filtering formula: ； In the formula, x 2 With y 2 represents the distance between the remaining pixels in the neighborhood and the center pixel in the neighborhood, respectively; c represents the standard deviation.

4. The method for enhancing details in low-light images according to claim 1, characterized in that, The following formula is used to perform a weighted fusion process on the brightness linear enhancement image and the detail extraction image: ； In the formula, I5 is the fused detail-enhanced image; w is the weight coefficient of the brightness linear enhancement image I3; and 1-w is the weight coefficient of the detail extraction image I4.

5. The method for enhancing details in a low-light image according to claim 4, characterized in that, The fusion coefficient is adjusted using the following formula: ； In the formula, δ(x,y) is the gray-level difference between adjacent pixels; P δ (x,y) represents the pixel distribution probability with a gray-level difference of δ between adjacent pixels.

6. A method for enhancing details in a low-light image according to any one of claims 1-5, characterized in that, The linear mapping coefficients are expressed as: ； In the formula, I0 is the original low-light RGB image; I5 is the fused image with enhanced details. The mapping relationship is represented as follows: ； In the formula, [R0,G0,B0] are the RGB color channels of the original low-light image, and [R',G',B'] are the RGB color channels after color remapping.

7. A detail enhancement system for low-light images, characterized in that, include: The image acquisition module collects low-light image data to obtain the original low-light RGB image; The color space conversion module converts the original low-light RGB image from the RGB space to the YCbCr color space and separates the Y channel grayscale image representing the image brightness. The guided filtering module uses the Y-channel grayscale image as a gradient guided image and performs guided filtering on it to obtain a smooth and denoised image that retains image details. The linear enhancement module uses a sigmoid function with an adjustable gain factor to perform brightness stretching on the grayscale image of the Y channel, and then performs Gaussian filtering on the brightness-stretched image to obtain a linearly enhanced brightness image. The grayscale difference module performs grayscale difference processing on the smoothed and denoised image and the brightness linearly enhanced image to obtain a detail extraction image that contains only image details; The fusion module performs image fusion processing on the brightness linear enhancement image and the detail extraction image, and adjusts the fusion coefficient using a fitness coefficient based on the gray-level difference between adjacent pixels and their corresponding pixel distribution probabilities, to obtain a fused detail-enhanced image with both enhanced details and brightness. The color space remapping module uses a linear mapping method based on the color information of the original low-light RGB image to perform RGB color space remapping on the fused detail-enhanced image, thereby obtaining a detail-enhanced low-light RGB image. The adjustable Sigmoid function for the gain factor is: ； In the formula, Imax and Imin correspond to the maximum and minimum gray values ​​of the original image, respectively; I represents the pixel grayscale value of the input image; a represents the parameter controlling the intensity of the linear gradient; b represents the parameter controlling the range of the linear gradient intensity.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1-6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1-6.

Images